Cell, Vol. 5, 213-225,

June

1975,

Copyright01975

by MIT

Molecular Information in Developmental Genetics Carol Newlon and Gary Gussin Department of Zoology University of Iowa Iowa City, Iowa 52242 Benjamin Lewin Cell The MIT Press Cambridge, Massachusetts 02142

Two recent meetings on developmental genetics displayed a clear consensus on the basic questions that lie at the forefront of present research. What is a eucaryotic gene? How is genetic information arranged in DNA and how is it expressed? What is the structure of the chromosome at the molecular level? How is genetic development coded in the genome? How are morphogenetic patterns specified? These were the principal topics of discussion at the First Annual EMBO Symposium, on Developmental Genetics, held in Hirschhorn (Heidelberg) from April 30 to May 3, and at the Symposium on Selected Topics in Developmental Genetics, held at the University of Iowa in Iowa City from May l-2. At both meetings it was apparent that important advances have been made in determining the organization of the genetic material itself and in elucidating some of the systems responsible for establishing positional information in development. Sequence

Information

Nucleosome Subunit Structure for Chromatin One of the most striking advances achieved in the past few months is the determination of a substructure for chromatin. The first suggestion for a simple repeating unit was provided by the electron microscopic observations of Olins and Olins (1974) that rat or chicken liver chromatin consists of “nu” bodies, particles appearing as beads on a string of DNA. Extension of these observations was described at the EMBO meeting by Chambon, who has used electron microscopy to visualize “nucleosomes” in HI-depleted chromatin (which is more readily dispersed for electron microscopy than chromatin containing Hl), in chromatin lysed directly from nuclei onto a grid, and in preparations reconstituted from DNA and all four of the histones H2A, H26, H3, H4. All these preparations contain tightly packed beads of diameter about 125 Aconnetted by a thread of 18-30 Athat represents free DNA. The nucleosome subunits contain equal amounts of histones H2A, H2B, H3, H4 and about 200 base pairs of DNA; histone Hl is irrelevant to their structure and formation (see Oudet, GrossBellard, and Chambon, 1975).

Meeting

Review

X-ray diffraction of chromatin produces rings at about 110, 55, 37,27, and 22 A. One model for the basic repeating unit which these rings suggest was constructed by Pardon and Wilkins (1972) based upon the concept that all the scatter rings were caused by DNA; this postulated a supercoiled fibre of diameter 1 OOA and pitch 11 O-l 20 A. At the meeting, Bradbury described experiments using neutron scattering which suggest an alternative interpretation. When the scatter from biological macromolecules is contrast-matched by varying the proportion of Hz0 to D20, histone is matched at about 37% D20 and DNA at about 63% D20. With chromatin, the 110 A band disappears and the 37 A is much reduced in the region from 30-40% D20; the 55 A and 27 A bands disappear at 60-70% D20. The 110 A ring thus appears to be due to protein scatter; the obvious interpretation is that it represents the internucleosome repeat. One speculation about the 37 A ring is that it may be due to a regular arrangement of proteins within the particle. The peaks at 55 A and 27 A which do appear to be due to DNA may reflect the organization of nucleic acid within the particle; one model consistent with these data is for DNA to be wound around the outside of the particle in 1.5-2 turns (see Baldwin, Boseley, Bradbury, and Ibel, 1975). A model for a basic repeating unit constructed of a histone octamer with the composition H2A2 H2B2 H32 H42 was originally proposed by Kornberg (1974) upon the basis of the ability of histones to aggregate in vitro studied by Kornberg and Thomas (1974). Kornberg described substantial evidence in support of this octameric structure. When chromatin is treated at pH 8 with the cross linking agent dimethylsuberimidate, a series of bands are obtained on SDS gels, representing histone monomers, dimers, trimers, up to much higher order structures. These multiple bands contain only the four histone classes H2A, H2B, H3, H4; two further bands alone contain Hl since these are missing from Hl-depleted chromatin treated with the agent. This regular spacing indicates that the histones are organized in a simple repeating unit; the octamer, with a size of about 110,000 daltons, corresponds well with the size predicted for a subunit containing two copies of each of the four histones (HI excluded). When the pH at which the crosslinking was performed was increased to 9, the bands corresponding to sizes beyond the octamer were absent, except for two weak bands that would correspond to a 16-mer and a 24-mer. Perhaps the increase in pH increases the net negative charge on chromatin, so that increased repulsion drives the nucleosomes apart, inhibiting crosslinking between particles.

Cell 214

These results thus clearly suggest that the basic protein subunit of chromatin is an octamer of H2A2 H2B2 H32 H4* (see Thomas and Kornberg, 1975). By reducing the concentration of chromatin, the octamer can be dissociated into a hexamer of H2A H2B H32 H42 and a dimer of H2A-H2B. The kinetics of crosslinking and analysis of the crosslinked dimers are consistent with more than one model for the organization of the nucleosome, but one general suggestion that appears attractive is that there may be a “core” of a tetramer of H32-H42 associated with two dimers of H2A-H2B. Digestion of nuclei or isolated chromatin with endogenous enzymes or with micrococcal or staphylococcal nucleases produces a series of bands of DNA on acrylamide gels; these represent simple multiples of a monomer whose size varies in reports from different laboratories from about 140 to about 200 base pairs. Noll (1974) obtained fragments of 205, 405, and 605 base pairs; and subsequent results in the Cambridge laboratory, reported at the meeting, suggest that smaller units may be derived by exonucleolytic attack on the ends of the fragments released by endonucleolytic action. Nibbling away of the monomer appears to proceed until a length of about 140 base pairs is reached, when resistance to further loss from the ends appears to be encountered. This suggests that some 200 base pairs of DNA are contained in a nucleosome and their arrangement on the particle allows a certain length at the ends to be readily digested by exonucleases. When DNAase I is used to treat chromatin, a repeating pattern different from that observed with other nucleases is seen. Kornberg reported that when the duplex DNA product is analyzed, no pattern of repeating units is found; but when denatured DNA is run on gels, bands are obtained at intervals of 10 bases. One model that might explain this result in terms of a structure for the nucleosome is to suppose that the DNA has “kinks” at 10 (or possibly 20) base pair intervals; it is possible to build a model in which there is a loss of stacking interaction, generating such a kink, which involves only two adjacent base pairs. One controversial point concerns the relationship between nucleosomes. Are they adjacent in vivo or may there be free DNA between them? Experimental evidence has been adduced in favor of the idea that free DNA may exist in the cell by Griffith (1975), who reported that the “minichromosome” of SV40 consists of 21 nucleosomes, separated by an average length of free DNA of 140 A. On the other hand, in reconstitution experiments Chambon has found that the number of nucleosomes into which SV40 DNA is condensed depends upon the condi-

tions, and it is possible to obtain a circular structure consisting of 25 nucleosomes which apparently contain all the SV40 DNA (see Germond et al., 1975). Some evidence on how the nucleosomes are organized into higher order structures in chromatin was presented by Bram, who by using neutron scattering at increased Bragg spacings has obtained a series of reflections at 400, 200, and 140 A. This would be consistent with a supercoil of pitch 500 Aand radius about 130 A, which could be accommodated by eight touching 100 Aspheres per turn (see Bram, 1975). By using freeze etching to visualize chromatin, Bram also has been able to obtain the pattern of a series of adjacent beads, lacking any interconnecting DNA. The subunits varied in diameter from 70-90 A, with an appearance more cylindrical than spherical. Some electron micrographs of chromatin suggest packing of nucleosomes into higher order structures; one striking photograph shown by Chambon represented a “zig-zag” of nucleosomes, but its significance is not yet clear. It seems likely that treating the nucleosome as a deoxyribonucleoprotein particle may lead to definition of its structure. Presumably this will reveal why the sequences of the histones in it are well conserved in evolution. The role of Hl in linking nucleosomes together is not yet understood. The apparent lack of any sequence specificity in the nucleic acid of the nucleosome leaves unanswered the question of how chromosomes are constructed with such precision so that, for example, exact crossing over can occur at meiosis. And a functional question concerns how DNA is recognized for transcription and also the related issue of what happens to the nucleosome at replication. The proposed location of DNA on the outside of the nucleosome may be significant in answering these questions. If all eucaryotic chromosomes are constructed of nucleosomes, the differences between them must reside in the nonhistone proteins. No new insights into the role of these proteins were offered, but Elgin reported the development of a technique which is potentially very valuable in allowing these molecules to be studied. Fluorescent antibodies prepared against total nonhistone proteins of D. melanogaster have been used to visualize both metaphase chromosomes and the polytene chromosomes of salivary glands. An antibody prepared against one gel band of nonhistone proteins (which of course may contain several species of protein) also works well in visualizing the polytene chromosomes. This technique thus offers the prospect of examining the distribution of different nonhistone proteins along the chromosome and of determining their tissue and species specificities.

Molecular 215

Information

in Developmental

Genetics

One Gene-One Chromomere For several years, a major problem has been that of reconciling the enormous haploid DNA content of eucaryotes with the relatively small numbers of complementation groups that can be identified genetically, particularly in Drosophila (see Bishop, 1974). Judd and Young (1973) have presented evidence consistent with the notion that each of approximately 5000 chromomeres corresponds to a single complementation group; yet the average chromomere contains about 30,000 base pairs, approximately 30 times the number needed to code for an average sized polypeptide chain. Judd himself has continued his exhaustive analysis of the zeste-white (3Al-3C2) region of the Drosophila X chromosome in an attempt to uncover new complementation groups. His original enunciation of the “one gene-one chromomere” hypothesis (Judd, Shen, and Kaufman, 1972) was based on analyses of complementation between lethal mutations spanning 15-16 bands in this region. Since then, he has attempted to demonstrate the existence of nonlethal mutants in new complementation groups within the same region. Two such complementation groups containing female sterile mutations have been described previously (Judd and Young, 1973). At the Iowa meeting, Judd reported the identification of two more nonlethal complementation groups, whose function affects circadian rhythms. Mutants labeled per produce an altered circadian rhythm period, and complement all the lethals in the region. The second function is not identified by point mutation, but by the fact that two overlapping deletions, J4 and D8, neither of which produce altered phenotypes in heterozygotes with wild-type, produce long periodicity in the J4/D8 heterozygote. Exhaustive attempts to uncover new genes in this region have thus met with limited success-the overall ratio of complementation groups to known chromomeres is still close to one. If the number of complementation groups yet to be uncovered in Drosophila is insignificant in terms of the amount of DNA the organism has at its disposal, what other functions for the DNA can be envisaged? Some loci in Drosophila may actually be control loci that are cis-acting and therefore not distinguishable by complementation from the structural genes whose activity they regulate. Judd has carefully examined two mutations, wch and WSP, which behave differently from mutations at seven other sites within the white locus. Mutation at the seven sites that are thought to be part of the w+ structural gene do not affect the expression of the zeste

phenotype

a deletion

in z$

of the white

females; locus

however,

or mutations

either at wch

and ww suppress the zeste phenotype in w+/w heterozygotes. A second effect of wch and WP, which also suggests that the region of the genome affected may serve a regulatory function, is observed in duplication mutants; a z w+ male, containing no duplication, is phenotypically wild-type, but a z w+ w+ male, containing a duplication of the white region on the X chromosome, is phenotypically zeste. The duplication need only contain the region defined by the two “regulatory” alleles to permit expression of the zeste phenotype in the male. Another system that has been studied in some detail in Drosophila is the rosy locus, which Chovnick reviewed at the EMBO meeting. Mutants that are rosy lack XDH entirely; to map mutations affecting the XDH protein, Chovnick therefore turned to isoalleles that alter its electrophoretic mobility without changing its enzyme activity. After deriving rosy mutants from each of five ry+ alleles, crossing two of these strains allows selection for ry+ recombinants, which can then be examined for electrophoretie mobility. Thus the ry mutant sites are selected markers and the sites responsible for electrophoretie mobility are unselected markers. The sites controlling electrophoretic mobility mapped in two clusters, one at each end of the region identified by rosy mutations. By using a cis/trans test, it was possible to show that both sites fall within the same protein coding element (see Gelbart, McCarron, Pandey, and Chovnick, 1974). A revision of the genetic map for rosy suggests that the outlying mutant sites are located 5 x 10-l map units apart. The enzyme of XDH comprises two identical subunits of 130,000 daltons each, representing a coding length of about 3600 base pairs. Thus one crossover unit corresponds to 720,000 base pairs. This is not very different from the value calculated on the basis that the Drosophila genome of 1.4 x IO* base pairs equals 275 crossover units and lends some confidence to the idea that recombination analysis represents distances along the chromosome. One of the isoallelic electrophoretic variants of the rosy locus produces a much more intense band of XDH activity than wild-type on gels. Titration experiments with anti-XDH serum showed that more is required to inactivate the extract of the heavy band than of the (wild-type) light band type, consistent with the idea that the heavy band results from increased production of enzyme molecules. In a preparation of heavy band enzyme, V,,, is increased 4 fold while the K, remains similar to wildtype. Although these results formally do not exclude the possibility that this is a structural variant of the enzyme, they fit well with the idea that the mutant causes a 4 fold overproduction. The site responsi-

Cdl 216

ble for heavy band production can be separated by recombination from the site responsible for the electrophoretic variation of its strain; the ryh site maps to the left of all known ry structural sites. It is c/s dominant and appears to increase production only of the electrophoretic variant carried on its own chromosome. These characteristics clearly are those expected of a control mutation. The rp site maps some 7 x 1 O-3 units to the left of the leftmost known ry structural mutant site. A limit upon the distance for which the rosy control and structural elements may extend to the left is placed by the mapping of the next leftward known mutant locus only 5 x IO-3 units away from rp. An obvious question is how far the rosy locus may extend to the right. Chovnick’s bias to the left in looking at mutations, prompted by the isolation of the interesting heavy variant, means that as yet no data are available. However, extension of mapping to the right may allow the proportions of structural and control elements constituting rosy to be determined. The rosy locus lies in a region of five bands on the polytene third chromosome, and since these all are about average in size, the band containing rosy presumably has about 10 times more DNA than codes for protein. Genetic analysis at this locus therefore offers some hope of defining a gene-band relationship. Results of the sort that have been obtained with white and with rosy suggest regulatory functions for two particular regions. The rarity of mutations that can be characterized in this manner means that no general answer is at hand, however, for the question of how extensive regulatory regions may be and how much DNA in the organism is devoted to control functions. Arrangement of Sequences in DNA It is now clear that the short interspersion pattern first described in Xenopus and in sea urchin is typical of the DNAs from many species: in these cases, a large proportion of the genome consists of short (about 300 base pair) lengths of moderately repetitive DNA adjacent to longer (about 1000 base pair) lengths of nonrepetitive DNA. As Davidson emphasized in his talk at the EMBO meeting, in sea urchin this fraction includes nonrepetitive sequences corresponding to the messengers of the gastrula, making very attractive the hypothesis that short period interspersion includes the alternation of control elements with structural genes (see Davidson et al., 1975). A much longer interspersion period has been observed in Drosophila, with repetitive elements of 6000 nucleotides alternating with nonrepetitive elements of 13,000 nucleotides, but Davidson notes that there is a component of about 15% of the Drosophila genome which may have the short period

interspersion pattern. In view of the small proportion of the genome apparently needed to code for protein, this might be sufficient to represent the structural genes. Drosophila is an attractive organism for investigating genome organization; since the bands of polytene chromosomes visibly identify what appear to be functional units, if it proves possible to determine the organization of DNA in a band, some correlation between genome organization and function may emerge. At the Iowa meeting, Thomas reported his recent analysis of the Drosophila genome with restriction enzymes; this clearly excludes any model for tandem repeat of genes within a chromomere, at one time apparently a mechanism which might explain the large amount of DNA. In these experiments individual restriction endonucleases were used to cleave DNA of Drosophila embryos and various tissue culture lines. If eucaryotic DNA were arranged in tandemly repeated sequences, then tandem sequences that do not contain a restriction site should all be “spared”; treatment with a restriction enzyme should yield long segments of uncleaved DNA. On the other hand, tandem sequences containing a restriction site should all be cleaved, producing fragments of calculated average length 1000-4000 base pairs, depending on the restriction enzyme used. Hamer and Thomas (1975) found that 5-10% of the Drosophila genome was spared by each of three enzymes (Eco RI, Hae, Hinll) and spared DNA appeared in fragments about 30,000 base pairs in length; however, most of this spared DNA appeared to be satellite DNA, based on in situ hybridization studies, isopycnic banding in CsCI, and renaturation kinetics. Data obtained in these experiments are consistent with either of two hypotheses: that not more than 10% of Drosophila DNA can exist in tandemly-repeated sequences 500-1500 base pairs in length; or there could be tandem repeats of length at least 4000 base pairs in every chromomere, which is still much more information than a structural gene is expected to contain. Cot analysis and digestion of DNAs with restriction endonucleases have been important in placing limits on the size and distribution of repetitive and nonrepetitive DNA sequences. Nevertheless, specific information about actual relationships between repetitive and nonrepetitive sequences and between regulatory and nonregulatory regions of the genome requires the ability to examine defined DNA segments. The tremendous importance of E. coli plasmids containing integrated eucaryotic DNA in such analyses was exemplified by Hogness’s discussion of his group’s progress in constructing plasmids containing integrated segments of the Drosophila genome (see Wensink et al., 1974;

Molecular 217

information

in Developmental

Genetics

Glover et al., 1975). The ways in which these plasmids can be used to examine chromosome structure are numerous, and studies of the first few plasmids already have provided significant information. Six of the plasmids isolated by Hogness’s group have been shown to hybridize in situ to single chromomeric regions of salivary gland chromosomes. From the kinetics of renaturation of each of these DNAs with total Drosophila DNA, they can be represented no more than twice in the entire haploid genome. Since the Drosophila sequences contained in the hybrid plasmids range in size from 8,000-18,000 base pairs, their possible uniqueness is clearly inconsistent with any tandem repeat model. However, studies of these plasmids also fail to support certain aspects of the Britten-Davidson model. All six plasmids contain only nonrepetitive sequences, a result that would not be consistent with short period interspersion. However, Hogness was careful to point out that short sequences repeated up to 10 times in the Drosophila genome would have been undetected in his experiments. The real power of the plasmids in revealing the nature of the organization of DNA in chromomeres can be seen from studies on plasmids which contain sequences that are repeated of the order of 100 times in the Drosophila genome. One such plasmid, pDM27, which hybridizes to the chromocenter and at least 70 chromomere regions of salivary gland chromosomes, is being used for two exciting experiments. The Britten-Davidson model suggests that structural genes whose products serve more than one function in the cell should be adjacent to more than one regulatory gene, and that individual regulatory element should be adjacent to all structural genes important in the process that they control. The in situ hybridization pattern of pDM27 suggests that it may contain at least one such putative regulatory sequence. To determine how many different reiterated (putative regulatory) sequences it contains, Hogness’s group is recloning sub-fragments of pDM27 in new plasmids and determining the in situ hybridization pattern of the sub-fragments. If pDM27 does contain a battery of regulatory sequences, then at least some subclones of pDM27 should hybridize to a subset of the chromomeric regions recognized by the parent plasmid. The second experiment of interest is to isolate all the sequences adjacent to the reiterated sequences in pDM27. Hogness has devised an elegant scheme for screening hybrid plasmids in vitro for the presence of particular sequences complementary to specifically chosen RNA species. This technique involves hybridization of radioactive RNA to denatured DNA contained in bacterial colonies grown on millipore filters. Using this method, Hogness’s group has now isolated about 30 hybrid

plasmids containing sequences pDM27, but which are different

in common from pDM27.

with

Processing of Transcripts The importance of defining the nuclear precursors for cytoplasmic messengers was emphasized by Davidson’s talk at the EMBO meeting in which he noted that in the sea urchin the complexity of nuclear RNA is some IO fold greater than that of cytoplasmic mRNA. Whatever selection of sequences for nucleocytoplasmic transport takes place is clearly crucial in understanding the expression of genetic information in eucaryotic cells. In reviewing the complexity of messenger RNA populations present in a variety of organisms, Davidson noted that (with the possible exception of Drosophila) a common range is lo-20 x 106 nucleotides, that is, about 5,000-10,000 genes of 2000 nucleotides each; this contrasts with wide variations in haploid genome size. Some progress towards demonstrating a precursor-product relationship for specific RNA species was reported by Darnell at the Iowa meeting from studies of the processing of adenovirus 2 “late” mRNAs. Using Eco RI restriction fragments of viral DNA as hybridization probes for specific sequences in isolated nuclear RNA, Darnell’s group has shown that poly(A)-containing viral messengers are synthesized as part of much longer RNA chains initiated in the A fragment (left half of the molecule) and are then processed very rapidly. Thus even after a 2 min pulse, RNA species that eventually appear in mRNA are already cleaved from the initial transcription product and contain no sequences able to hybridize to the A fragment. This very fast processing time suggests that it will be extremely difficult to isolate large quantities of uncleaved hnRNA to use in biochemical studies without first finding mutations or drugs that affect processing. Very elegant evidence that the primary transcription product is very large was reported at the Iowa meeting by Laird. He has adapted the technique of Miller and Beatty (1969) to the electron microscopic visualization of transcription of chromatin from Drosophila embryos. Two kinds of transcription units are found: those typical of ribosomal cistrons in other systems, with a gradient of densely clustered ribonucleoprotein fibrils separated by nontranscribed regions in the familiar Christmas tree array; and those characterized by much less densely spaced fibrils, which presumably represent nonribosomal transcription products. Since the density of presumed messenger fibrils is only about 1-6 fibrils per p of chromatin, compared with 30-40 fibrils per fi for ribosomal transcription units, the frequency of transcription initiation may be much lower for messenger species than for rRNAs.

Cell 218

Twelve of the putative messenger transcription units have been analyzed by mapping the lengths of individual fibrils against their position of attachment to the chromatin. Seven of the units exhibit a simple pattern in that the lengths and distribution of fibrils suggest that transcription proceeds from a unique origin to a unique terminus. The average length of these transcription units (~2 p) corresponds to about 6 (J of DNA, which is approximately the same as the average length of DNA in a chromomere. For the remaining five units, a more complex pattern was observed, with each unit appearing to contain several adjacent gradients of length that could be extrapolated to a fixed point in the chromatin. These patterns can be interpreted in either of two ways. They may represent adjacent transcription units with overlapping origins and termini, or more likely they result from specific processing of very long transcription products during synthesis. If the latter interpretation is in fact correct, it would provide an explanation of the observation that the average size of hnRNA in Drosophila is only about 7000 bases even though transcription units may contain an average of 25,000 bases. In this case, giant hnRNAs may provide a preliminary explanation of the use to which excess DNA in chromomeres is put. The problem will then be to determine the function of excess RNA rather than excess DNA. DNAs Containing Tandem Reiterated Sequences Although tandemly repeated sequences may not be a major component of chromomeric DNA, sequences specifying t-RNA are tandemly repetitious (Brown and Sugimoto, 1973; Wellauer and Dawid, 1973). In Xenopus, the nucleolar organizer region of the chromosome contains 500 copies of the ribosomal genes; adjacent copies are separated from each other by nontranscribed spacer DNA. Heteroduplex mapping of this DNA has shown that the transcribed sequences are all very similar, if not identical, while the nontranscribed spacers are frequently different in length and sequence. Dawid reported on progress he and Wellauer (in collaboration with Brown and Reeder) have made in understanding the differences in the spacers. These studies were facilitated by the fact that EcoRl restriction endonuclease cuts the ribosomal DNA into pieces containing most of the transcribed region, which are all 5000 base pairs in length, and pieces containing mostly the nontranscribed spacer region, which range from 5,900-10,000 base pairs in length. Several of these “spacer” regions have been inserted into E. coli plasmids and cloned (see Morrow et al., 1974). The pattern of single stranded loops obtained from heteroduplex mapping of cloned spacers revealed that the spacer contains

internal repeats of a sequence less than 50 base pairs long, which is consistent with hyper-sharp melting profiles of spacer DNA. These internal spacer repetitions are organized into four longer regions, A, B, C, and X. Dawid reported that regions B and X seem to be constant in length, while regions A and C are expanded in spacers of greater length. Heteroduplex mapping of spacer regions in DNA from individual frogs revealed that some frogs contain up to seven different spacers in their rDNA, and that different spacers are most often found adjacent to each other in chromosomal rDNA. In fact, the frequency with which spacers of identical length are adjacent suggests that there must be a random “scrambling” of tandemly repeated rDNA segments (containing 40s unprocessed rRNA genes plus spacer sequences). In contrast to chromosomal rDNA, the repeating units in amplified rDNA are not scrambled; adjacent spacers are always the same. This result is expected if amplification proceeds by a rolling circle mechanism and suggests that rolling circles originate from the excision of a single ribosomal repeat. Thus rolling circles several units in circumference probably arise not from excision of several tandem repeats but from some later intramolecular recombination event. Nontranscribed satellite DNAs in Drosophila seem to resemble the nontranscribed spacer of rDNA in several respects. Sequencing studies (Gall, Cohen, and Atherton, 1973) indicate that satellite DNAs consist of repeating nucleotide sequences 6-12 nucleotides long. However, the repeats are not exact and treatment of Drosophila DNA with restriction endonucleases in Thomas’s lab has revealed much longer periodicities in satellite DNA. Among the spared segments is a modular series of multiples of a 365 base pair repeat. The segments are contained in satellite DNA with density 1.688 g/cc (Manteuil, Hamer, and Thomas, 1975). These longer periodicities may perhaps be analogous to the longer regions revealed by heteroduplex mapping on nontranscribed rDNA spacer. The significance of these repeats is unknown as is the function of these nontranscribed DNAs. Positional

information

Control of Phage Development Studies of phage assembly demonstrate that individual proteins and phage structures may exist in the cell without assembling. In these situations, the appearance or activation of a single gene product may suffice to trigger the entire assembly process. This is exemplified dramatically by King’s studies of the assembly of the tail of phage T4, reported at the Iowa meeting. At least 21 known T4 gene

Molecular 219

information

in Developmental

Genetics

products are needed for assembly of the tail, 18 of which are found in the complete Structure. Cells infected with mutants defective in these 21 genes accumulate unassembled proteins and incomplete structures which have been characterized according to their size and morphology, the proteins they contain, and their ability to serve as precursors for assembly in vitro (see Kikuchi and King, 197.5). Several important principles emerge from these studies. First, the interactions of gene products occur in strict sequence even through the proteins themselves exist simultaneously in solution throughout the period of morphogenesis. This kind of sequential assembly has two important implications: the first step in the morphogenetic pathway becomes the “trigger” for succeeding steps, in that the presence of the first protein in the pathway is required for initiation of morphogenesis; it is almost impossible to assemble defective tail structures, since any structure that incorporates a defective protein should be incompetent to take part in subsequent steps in the assembly sequence. A second principle is that of sub-assembly which has been enunciated previously by Wood, King, and Edgar. Six minor proteins and catalytic gene products interact to form the central “hub” of the baseplate; independently, seven other proteins interact sequentially to form a 145 “arm” complex, which appears to be a one sixth “wedge” of the complete hexagonal structure of the baseplate. Six such “arm” structures aggregate around the 22s hub to form the baseplate, to which further gene products must add to activate the baseplate for tail tube polymerization. The independent assembly of the “hub” and the six “arms” of the baseplate are a further insurance that the final structure will be able to function normally, since defective sub-structures can be rejected at the convergence of the two pathways. Finally, King conjectured that constraints on the assembly process are not designed to reject defective sub-structures so much as to ensure that the normal finished product can function effectively. He stressed the dynamic nature of the finished structure, and the complexity of interactions between assembled proteins, which may necessitate a carefully planned sequence of assembly steps. The importance of these findings was demonstrated when, later in the meeting, King used sequential assembly as the basis for a model to explain certain cases of embryological determination. A different kind of developmental system-the cell infected by bacteriophage h-was discussed by Echols. One of the developmental choices available to X consists primarily in the selection of one of two mutually exclusive transcriptional states characteristic of the lytic response or of lysogeny. Echols

outlined detailed analyses of genes involved in the regulation of h repressor synthesis, studies that have been carried out in several laboratories. An interesting feature of the system is that the proteins coded by two genes, cl (which specifies repressor itself) and cro (or tof), are each the product of an operon repressed by the other. Thus in a lysogen only the repressor is present, and only the cl operon is transcribed, while in a lytically-infected cell, the cm gene product eventually turns off transcription of cl, so that transcription of genes required for phage replication can proceed at a maximal rate. An impressive feature of the lambda system is that regulatory gene products have multiple functions in both the lytic and lysogenic pathways, and the regulatory sites can be recognized by more than one effector protein. Early Embryogenesis in Drosophila Studies of simple systems like the bacteriophages are clearly important both for model building and for understanding the nature of interactions that potentially can occur at the molecular level. Although the complexity of embryogenesis of multicellular eucaryotic organisms has been a barrier to such a thorough step-by-step study, the power of genetic analysis of complex developmental processes is impressive. At this level, Drosophila provides the opportunity to combine genetic and developmental techniques to study the nature of “irreversible” commitments that occur during differentiation. Results reported at both meetings show progress in understanding early development, the processes involved in determination of imaginal discs, and the establishment of compartments in disc development. Before the blastoderm stage in Drosophila, nuclei migrate to form a layer at the surface of the embryo; the first nuclei reach the posterior end and form pole cells containing characteristic polar granules. These give rise to the primordial germ cells. Can cytoplasmic determinants be identified in the polar plasm? At the EMBO meeting, lllmensee described his experiments on injecting polar plasm into the anterior or ventral position of early cleavage recipients; this causes the appearance of pole cells. To detect the formation of functional cells, lllmensee performed two serial transfers: first polar plasm was transferred to the anterior or mid-ventral position of an early cleavage recipient; then the recipient was allowed to develop into a blastoderm and cells from these positions were transferred into another embryo, at the posterior position among the pole cells. By using different genetic markers in each of the three sets of embryos, it is possible to distinguish the genetic source of cells in the fly. The presence of germ line mosaics showed that the polar

Cell 220

plasm could induce formation of pole cells in the anterior position of the first recipient. This provides a striking demonstration of the presence of specific cytoplasmic determinants in the posterior region of the plasm (see lllmensee and Mahowald, 1974, 1975). When does cytoplasm acquire the specificity to determine germ ceils? lllmensee investigated this question by using as donors oocytes of different stages; stages 13 and 14 give the same result, with 4-5% germ line mosaics. Stages 10, 11, 12 give no germ line mosaics and cause the formation of small premature polar bodies that do not aggregate properly. Are the determinants species specific? Drosophilaimmigransformspolargranulesthatarelarger than those of D. melanogaster and which aggregate in a characteristic manner to form large clusters. The cytoplasm of D. immigrans induces the immigrans type of structure upon injection into melanogaster and can cause production of active primordial germ cells. In the next generation, however, all the granules are those characteristic of melanogaster, so the effect is not inherited. Another question that can be asked by transfer experiments is what degree of precision exists in the determination of structures in the early embryo. lllmensee has transferred cells from one blastoderm to a different position in another blastoderm and then investigated what structures are formed. When cells are transferred to a homologous position, integrated structures are formed which display genetic markers characteristic of both donor and recipient cells. When cells are transferred to a nonhomologous position, however, segregated structures are formed: the donor cells independently form structures that would have been developed had they remained in their original position. One interesting feature of these transfer experiments is that in the flies studied so far, a single transferred blastoderm cell gives rise only to adult cell types that can be derived from a single type of imaginal disc; that is, at this stage there appears already to be considerable restriction on the potential development of the cells. A technique used in several experiments with Drosophila to investigate the timing of events is the induction of somatic crossing over by irradiation. Using a fly of heterozygous genotype + /m, X irradiation may induce a crossover event at mitosis; if this occurs between the site of heterozygosity and the centromere, the progeny cells are both homozygous, m/m and + /+. The descendents of the m/m cell then show the phenotype of the recessive m mutation and so can be recognized as a clone. Gehring has used this technique to determine the precision of determination of blastoderm nuclei. After inducing somatic crossing over in the blasto-

derm, adults are examined to see whether clonal descendents can be found in structures derived from more than a single type of disc. The descendents of one blastoderm nucleus were found in both wing and second leg; the frequency of such an occurrence is about the same as that of joint descendents of cells in the different extremities of one disc. With the antennalieye disc, irradiation at 3 or 7 hr produces about 10% clones restricted to the antenna, about 60% restricted to the head, and about 20% overlapping. No distinction can be seen between the larval cells of eye and antenna1 discs, even though they are determined to give rise to quite different adult structures. When the antennal/eye disc is stained for the enzyme aldehyde oxidase, however, the antennal part shows a blue stain and the eye part is not stained, with a sharp boundary between them. There is therefore a clear difference, which develops during the second larval instar, between these two discs in at least this enzyme activity. lmaginal Disc Determination How is the ability to maintain a particular determined state genetically controlled? Since cell commitments are initiated by egg cytoplasm in contact with the nucleus, at least some mutations that alter determination should be manifest as maternal effect mutations. One such mutant, “tumorous head,” has been identified; other mutations whose effects are not maternal include aristapedia and antennapedia (see Postlethwait and Schneiderman, 1971). The existence of these mutants suggests that the gene products affected must be present at least once during larval development if reprogramming is to be avoided. Postlethwait has used these homoeotic mutants (mutants that exhibit altered determination) to investigate the nature of the requirement for genetic information in the maintenance of a particular determined state. Through X-ray induced somatic crossing-over, it is possible to produce clones of cells that have lost a gene necessary to avoid reprogramming; for example, a specially constructed strain that is homozygous for aristapedia (ssa) on the third chromosome, but which also contains an insertion of the ss+ allele on one X chromosome, is phenotypically wild-type. Somatic crossing over and repulsion of chromatids in mitosis produce new assortments of chromosomes in the two daughter cells, frequently producing a cell that has lost the ss+ duplication and has thus become genetically SS~/SSQ. Linkage with the X chromosomal marker y (yellow) results in a situation in which wild-type ssf cells remain y+ (black) and produce normal aristae structures, while cells that have lost ss+ are yellow (y/y) and produce leg structures. By altering the time of X-ray treatment, it is possible to examine

Molecular 221

Information

in Developmental

Genetics

the time at which normal gene activity is no longer necessary to maintain the determined state. From the results that Postlethwait reported at the Iowa meeting, it can be concluded that activity of the ss+ gene at least one time until the third instar larval stage is required to maintain the commitment of antenna1 discs to produce antennae. Further, in the mosaics that are produced, all yellow cells produce leg parts, and all black cells produce antennae, suggesting that the two cell types produced by somatic crossing over differentiate autonomously; the required gene product only acts in the cell in which it is produced. And if larvae are X-rayed within 3 cell divisions of the onset of morphogenesis, y marker clones develop normal aristae. The effects of the wild-type gene thus can be maintained for at least three cell generations. The problem in such studies is in distinguishing between the initial commitment, or determination, and the ultimate fulfillment of that commitment. Are the genetic loci analyzed using mosaics actually required for the initial program, or do they merely produce building blocks necessary for the ultimate fulfillment of the program? How do cells that are adjacent in a disc recognize their position in a disc and ultimately differentiate into different sub-structures? Implantation of a segment of a disc into an adult results in regeneration of the mirror image of that segment. If somatic crossing-over is induced by X-rays prior to excision of the segment, the resulting mirror image is not exact. That is, marked clones do not necessarily occupy corresponding positions. Thus position within a disc is not strictly inherited, and duplicate copies can originate from a small number of cells. If half the original cells in a disc are killed at random with X-rays, the adult fly does not form amorphous structures, but forms recognizable multi-structural parts. For example, if wing discs are X-rayed: a majority of the adults form normal wings; some adults form duplicate structures with mirror image symmetry; all structures are either duplicated or missing: missing parts always occur at the boundary of the duplication, that is, at the point where the two mirror images meet; some triplications occur, in which case 2 planes of mirror image symmetry are observed, one between each pair of triplicated structures. Thus cells are not confused, for the survivors follow specific instructions about how to assume their position in the final structure. One model which may explain these results is to suppose that there is a conical gradient with its high point in the center of the disc, Cells can often proliferate to produce cells lower than themselves in the gradient. Kauffman (1973) has proposed a model to interpret the results of these kinds of studies. Based on the existence of determined states, frequencies of

transdetermination, the existence of certain classes of lethal mutants, and the general direction of transdetermination from certain states to others, Kauffman has suggested the existence of a binary “epigenetic” code underlying the process of imaginal disc determination. According to the model, the determined state of a particular disc depends on seveach of which can exist in eral binary “circuits,” only one of two alternate states. The relative stabilities of the two states of a particular circuit regulates the frequency of transdetermination, which may occur through a “switch” in the state of that particular circuit. For example, Kauffman has assigned the binary-coded state 11000 to genital disc, and 11010 to leg disc. Here, genital disc cells would transdetermine to leg cells by a switch in the fourth “circuit” from 0 to 1. Compartments in the Drosophila Wing Disc Induced crossing over has been used by GarciaBellido and by Lawrence to map compartments in the wing disc. When larvae that are heterozygous for the recessive mutant mwh (multiple wing hair) are irradiated, the wings of the adult flies contain patches of cells of the mwh/mwh type. Each patch is clonal: it represents the descendents of a single cell in which the crossover took place. The earlier the time of irradiation, the larger the patch, since the single cell that suffered mitotic recombination represents a larger proportion of the (smaller) number of precursor cells. A variation of this technique which makes the mutant patches easier to visualize, especially when they are derived from irradiation at later times and therefore would be small, is to utilize a minute mutation. Cells that are heterozygous for a recessive minute mutant grow more slowly than cells that are homozygous wildtype; by utilizing an appropriate minute, it is therefore possible to create a situation in which mitotic crossing over generates a cell that is marked by the homozygous mwh/mwh type and the -l-m/ l-m type; this should grow more rapidly than the + m/m cells and thus gives clones of increased size. A clone produced by very early irradiation may occupy any part of the wing. Irradiation at early times produces clones that may occupy irregular areas in either the posterior or anterior part of the wing, but which fail to cross a straight boundary located not far from one of the veins across the wing (see Garcia-Bellido, Ripoll, and Morata, 1973). This boundary therefore defines the posterior and anterior compartments of the wing. In the terminology proposed by Crick, the compartment is said to comprise a polyclone, where the polyclone is all the surviving descendents of a small group of founder cells, The polyclone is confined to a compartment, but the positions of cells within the compartment

Cdl 222

are not specified; for the fact that the clones identified by somatic recombination in different individuals may overlap within a compartment indicates that cell lineage in Drosophila is not strictly laid down but follows general rules. Irradiation at later times sees the establishment of new boundaries that cannot be crossed, separating dorsal from ventral. The dorsal/ventral decision appears to be made at the same time in both anterior and posterior compartments. These results thus are consistent with the idea that a binary decision is made at some time such that first the cells of the developing wing disc are divided into anterior/posterior types and later into ventral/dorsal types and then into other subcompartments. Questions that these results raise include how the boundary is formed and how it is established in such a straight line. One important implication of these results is that even when cell growth rates are greatly changed, the same overall size, shape, and division into compartments is maintained in the wing. The mechanism responsible must therefore be independent of cell growth rates. Results Lawrence reported at EMBO with the mutant engrailed (en) suggest that the en+ gene may act as a switch responsible for controlling the ability of cells to cross the anterior/posterior boundary. This mutant has pleiotropic effects, characterized by Garcia-Bellido and Santamaria (1972), and causes a transformation of the posterior part of the wing into anterior-like parts; this is seen in the development of a line of bristles along the posterior margin, bristle formation usually being a feature only of the anterior margin. So in this sense engrailed represents a homoeotic mutation. In order to follow mutant clones, Lawrence used a system in which larvae are heterozygous for own, en, minute. A somatic crossover generates pwn/pwn, en/en, + m/ + m cells which can be visualized by the pwn/pwn within the wing, by en/en at the margin, and which have a growth advantage over the other, + m/m cells. Clones of this genotype arising in the anterior compartment can fill up to the whole of the compartment but continue to respect the line dividing anterior from posterior. Clones arising in the posterior compartment, however, are able to cross the boundary and can fill up the whole wing. This is consistent with the idea that the engrailed gene usually is switched on in the cells of the posterior compartment and acts to prevent them from crossing the boundary line; the gene is not active in the anterior compartment. That is, en+ is the gene responsible for establishing the characteristics of the posterior compartment (see Morati and Lawrence, 1975). Engrailed produces a duplication in the anterior part of the leg and it is therefore possible that this same gene is used in a switch capacity in the leg as well as wing.

Use of somatic recombination allows the shapes and sizes of clones to be followed. Garcia-Bellido reported that in the anterior part of the wing, clones are elongated in parallel with the boundary. In the posterior part, however, they tend to run inward from the edge. The en/en clones, however, are elongated parallel to the edge in the posterior compartment. This suggests that one effect of en+ may be to control the orientation of cells after mitosis. Lawrence reported that small clones of engrailed cells that touch the posterior edge of the wing can cause a deformation in a larger area. The change results if and only if the clone touches the border and is dorsal. This suggests a pattern of nonautonomous development in which the shape and size of the wing may be controlled by the genotype of the cells along the edge. In this way the engrailed gene may control the shape of the posterior compartment. From a small number of genes of this type, it would clearly be possible to construct a binary system that specifies the development of compartments. Obviously these properties are compatible with the idea that en+ is a regulator gene; Garcia-Bellido suggested that it should be called a selector gene. At the EMBO Meeting, Sander reviewed an impressive array of evidence, depending upon work with a variety of insect eggs, which showed that gradients are present, revealed by regeneration or segmentation experiments which cause cells to recognize the loss of neighbors and to respond accordingly. These experiments argue strongly against the view that the egg consists of a preformed mosaic. Mammalian Developmental Systems Drosophila is an obvious choice for studying determination and differentiation, and it is much less obvious which mammalian systems are suitable for such studies. From the systems which were described at the EMBO meeting, however, it was clear that the same conceptual questions are being asked: what is the nature of the determining event and how is development executed? Progress with the mouse teratoma system was reviewed by Jacob. Teratocarcinomas are common only as testicular tumors in strain 129 or as ovarian tumors in strain LT2; teratocarcinomas in the testis of strain 129 may be established at about 12 days of embryonic development, the time when primordial germ cells arrive at the genital ridge. These tumors contain both nondifferentiated cells, the embryonal carcinoma cells, which are multipotent and retain a perfect mouse karyotype, and they also contain cells with a variety of differentiated phenotypes. There are therefore analogies between the cells of the early embryo and the embryonal carcinoma cells, but the ability to culture the teratoma

Molecular 223

Information

in Developmental

Genetics

system in vitro means that analysis is possible in a manner not possible with the whole organism. One object of this work is to obtain lines that are limited in their potential to differentiate; no such stable line has yet been obtained, however. One of the lines of embryonal carcinoma cells grows undifferentiated in culture so long as it is transferred every 4 days; when it is allowed to grow to confluence, differentiation occurs after 12-l 4 days, progressively leading to the appearance of several differentiated cell types. All attempts to make these cells differentiate without growth to confluence have failed. But if cells of such a nondifferentiated culture are centrifuged to form a small pellet, when the pellet is cultured it attaches to the surface and forms a colony. The period before the first morphologically differentiated cell appears is inversely related to the initial size of the colony. While it has not yet been possible to obtain cell types of restricted developmental potentials, this system is in many ways a promising mammalian counterpart to the imaginal disc system of Drosophila, in which it should be possible to study both determination and differentiation. The potential of the teratoma system for investigating problems that are difficult to study experimentally with the embryo is shown by the use to which it has been put in examining surface antigens. Somatic cells of strain 129 carry the H2b histocompatibility antigen; but this appears to be absent from the teratoma cells. A search for other antigens identified a component on teratoma cells that is present on normal morulae. Before or just after fertilization of the egg, there is a very low, barely detectable level of this antigen; the level increases until the 8 cell stage, when there is as much as on teratomas, and then declines as development progresses. The only other cell type found to carry the antigen is sperm (see Artzt, Bennett, and Jacob, 1974). Examination of human sperm showed the presence of the same antigen; it is found also on rat morula but not in chick (see But-Caron et al., 1974). It is attractive to speculate that this may be an antigen distinguishing mammalian sperm and very early embryos. When does the antigen appear on sperm? A fluorescent technique locates it on the postacrosomal membrane, precisely the region that fuses with the egg. All cells of the male germ line fluoresce; and in a mutant which lacks male germ cells there is no fluorescence. What genetic locus specifies the sperm antigen? Mutants at the T locus of the mouse form a series of recessive lethal% each of which acts at a specific stage of development. Since some T mutants have been implicated in cell surface defects, this locus is a good candidate to specify this surface antigen;

the earliest acting mutant is t’*, which acts at the morula stage. An indirect assay for the effect of t’* on antigen formation showed that this locus does indeed code for the antigen. When the fate of the tl* and H2 antigens was followed during teratoma differentiation in vitro, H2 appeared at 3 days, and the proportion of cells carrying it then increased; the t’* antigen declined as H2 increased. An examination of sperm for other antigens showed that H2 is absent; the antigen previously thought to be H2 is in fact IR. The antigens tl* and H2 may therefore be mutually exclusive. Myoblast differentiation provides another system that can be studied with some facility in vitro and progress with this was described by Yaffe. When myoblasts from the newborn rat are placed in tissue culture, they can fuse after about 50 hr to form multinucleated bodies that generate contractile crossstriated fibres. By maintaining cells in fetal-enriched medium, it is possible to prevent fusion; transfer to standard medium causes fusion. By collecting cells from fusion-promoting medium before fusion occurs, Yaffe has shown that the commitment to fusion can be reversed for up to l-2 hr before the event itself. It is possible to follow the appearance of enzymes and mRNA species during the course of differentiation. Actinomycin suppresses fusion, but if added at early times of culture causes an increase in myosin synthesis. By examining the ability of mRNA from these cells to produce myosin polypeptides in vitro when translated by wheat germ extract, Yaffe found that the message appears to be produced in the cell long (about 10 hr) before the protein is translated from it. He therefore suggested that translation of the myosin message may be controlled by a protein that is translated from a short lived mRNA which is susceptible to actinomycin inhibition; this would explain how actinomytin stimulates myosin synthesis at early times. Nongenetic Patterns The basic theme of Sonneborn’s address at the Iowa meeting was the diversity of levels at which control may be exerted in development. He stressed at the outset that natural selection selects for ends rather than means, so that there is no reason to expect a priori that the underlying mechanisms are uniform. The remainder of his talk was devoted to illustrating this proposition in a great diversity of systems, but with a consistent focus on the relationships of developmental mechanisms to the gene. The variety of relationships dealt with ranged from selective loss of genomes at the one extreme, to the hereditary maintenance of diverse spatial patterns in the presence of a constant genome at the other, with the more familiar mechanisms of transcriptional and translational control in between.

Cell 224

In certain examples, the choice to render a gene active or potentially active can be made only at a specific stage, with a stable inheritance thereafter of the state of activity or of potential activity. Examples of such determination of states of gene activity come from systems as diverse as the control of mating types in ciliates and the action of the “ovadeficient” gene in axolotls. Such fixation of a state of gene activity which is thereafter inherited is an analogue at the level of known gene activities to the inheritable state of cellular determination stressed by Postlethwait and Kauffman in their symposium talks. A dramatic example of differences of cell surface configuration nof caused by genie differences is the inversion of a ciliary meridian in Paramecium aurelia, first described by Beisson and Sonneborn (1965). Under certain special circumstances, paramecia may obtain grafts of one or more ciliary meridians that are rotated 180”. All structures associated with the inverted meridians are “upside down.” The inversion has been observed to be propagated for as many as 800 cell generations, despite the demonstrated absence of any relevant genie differences between cells that carry the inversion and ceils that do not. In no case has an inverted meridian been known to “reinvert” itself. These findings, coupled with ultrastructural observations on the mode of production of new ciliary units within ciliary meridians (Dippell, 1968), support the conclusion that the orientation of these structural elements in space is determined by the local intracellular environment (Sonneborn, 1970); genes supply the structural materials that make up the pattern, but do not determine the specific positions of new organelles. The self-perpetuating ciliary meridian inversions are not trivial in their effects on the cell, since the inverted regions do not participate normally in the development of larger structural features such as the cell anus and cell fission line, and furthermore manifest an inversion in direction of ciliary beat which results in abnormal swimming by the cell (Tamm, Sonneborn, and Dippell, 1974). Variations in cell organization such as these are just as readily subject to natural selection as are more conventional gene-controlled phenotypic variations. It is conceivable, though not proven, that self-perpetuating assemblies might exist in other organisms as well as ciliates. One system in which structural assembly is displayed to striking effect is the interaction of dermis and epidermis, which takes place at an early stage of development and is necessary for the production of hairs, feathers, and scales in mammals, birds, and reptiles. At the EM60 meeting, Sengel reviewed an extensive series of experiments on recombination of dermis and epidermis. Reconstitutions using

mouse dorsal and upper lip, chick dorsal and tarsometatarsal, and lizard dorsal and ventral regions suggest that the pattern is imposed by the dermis but that the type of structure formed is dictated by the epidermis. In heterospecific transplants, mouse dermis in combination with lizard or chick epidermis thus causes the production of scales or feather buds, respectively, but whose pattern (size and orientation) is that typical of the mouse. These scales and buds do not develop into mature structures. However, it is very striking that such heterospecific interactions can take place and can convey positional information of clearly identifiable origin. Sengel suggested that there are two types of interaction in the development of mature structures: first a heterospecific dermis can trigger a nonspecific induction which causes formation of epidermal prestructures, scales, feathers, hairs; but to construct a mature structure, a class specific induction is necessary from dermis of the same species. This idea is supported by experiments utilizing dermis from chick or mouse regions that do not form feathers or hairs. When mouse epidermis is recombined with chick dermis, the addition of cells from such a region of mouse dermis is sufficient to allow the formation of typical hair structures. Similarly, when chick epidermis is recombined with mouse dermis, only arrested feather buds can grow; addition of a few chick cells from a nonfeather forming dermis region allows formation of typical chick structures. This confirms the two step model: the heterospecific dermis can provide the first set of instructions; and the heterotopic cells can provide the second trigger. lntraclass recombinations between chick and duck showed that all important structural features of the feather depend on the dermis (those depending on the epidermis represent only minor cytological features), whereas the polarity of the feather depends entirely upon the epidermis. As Sengel concluded, the transfer of information between cell types is clearly established in these experiments, and it remains for the molecular biologist to investigate its nature. Construction of the Nematode Nervous System Striking progress has been made by Brenner in his study of the nervous system of the nematode C. elegans; at the EMBO meeting he discussed the dissection of its assembly. By making serial sections, Brenner has constructed what is a wiring diagram for the pattern of neuronal connections. A newly hatched nematode turns out to have only 15 neurones in its nervous system; new cells appear in the next developmental stages-that is, over the next 8-10 hr-and 54 new neurones are formed to bring the total number up to its mature value. (This

Molecular 225

Information

in Developmental

Genetics

is in contrast with previous ideas that the larval nematode contains all adult systems and needs simply only to grow.) By using Nomarski optics to follow neuronal development under the microscope over the next 3 days, Brenner has shown that the new neurones are derived from 12 precursor cells, 6 laterally arranged on each side of the organism. Each of these precursors shows precisely the same pattern of divisionthat is, produces the same “tree” of descendents. The divisions of each precursor generate six adult cells, arranged from posterior to anterior with successive divisions; these cells can be characterized either by their position or by the number of cell divisions needed to generate them. The only variation in the six trees on each side is that one cell dies in both the posterior and anterior trees and another cell dies at the posterior end. Since the posterior cell of each tree becomes a hypodermal cell, five neurones remain in each of the 12 trees, except for the posterior and anterior trees where programmed death reduces the number; thus a total of 54 new neurones is produced. These new cells insert themselves among the preexisting 15 juvenile cells, which fall into five sets of three cell types. (Little is known about the lineage of the juvenile cells, since it is difficult to follow their appearance in the very early stages of development.) This insertion results in an apparent variability in the construction of the nervous system from animal to animal. One crucial implication of these studies therefore is that the development of the nervous system cannot be deduced from the final pathway that is present in the adult; rather is it necessary to follow its assembly during development. How do these cells know what to become? Two models are to suppose that they compute their fate at each division (lineage computation) or from their neighbors (positional computation). One experimental way to resolve these models would be to change the positions of two cells and see what they become. Although this is not possible, a close approximation is to kill cells selectively during development and see what changes ensue in the remaining cells. Although there are at present no mutants available which have a neurone missing, there is a mutant which affects two specific cell types, which are very similar morphologically-one derived from the juvenile and one from the adult pathway; these cells fail to form synapses in the mutant. Therefore one gene at least may function because of the role of the cell, that is, independently of its lineage. From the results Brenner discussed at the meeting, it is clear that a considerable advance has been made towards the aim of defining completely the construction of the nervous system of the nematode;

and already ing clear.

the specificity

of its assembly

is becom-

We wish to thank Drs. D. R. Soll, J. D. Mohler, and J. Frankel for their help in the preparation of this manuscript, and also many others at both meetings who offered helpful comments.

Artzt, K., Bennett, USA 71, 811-814.

D., and Jacob,

Baldwin, J. P., Bosely, Nature 253, 245-249.

P. G., Bradbury,

Beisson, J., and Sonneborn, USA 53, 275-282. Bishop, Bram,

J. (1974). S. (1975).

Brown, Quant.

Nat. Acad.

Davidson, E. H., Hough, Cell 4, 217-238. R. V. (1968).

Cold Spring

Proc.

Nat. Acad.

A., and Santamaria,

Germond, J. E., Hirt, bon, P. (1975). Proc.

Griffith,

J. D. (1975).

Hamer,

0. H., and Thomas,

Science

Genetics

Spring

72, 87-104.

G. (1973).

Nature

J., and Chovnick,

New

A. (1974).

K., and Mahowald, 020.

Illmensee,

K., and Mahowald,

Judd, B. H., and Young, Quant. Siol. 38, 573-579.

A. P. (1975).

Science

Y., and King, J. (1975). Science

R., and Thomas,

Proc.

49, 243.

Nat. Acad.

Exp. Cell Res.,

Cold Spring

Kaufman,

in press.

Harbor

T. C. (1972).

Sci.

Symp. Genetics

787, 310-317. Submitted

for publication,

184, 868-871, J. 0. (1974).

S., Hamer,

D. H., and Thomas,

O., and Beatty,

B. R. (1969).

G., and Lawrence,

Chromosoma

A. P. (1974).

M. W. (1973).

S. A. (1973).

D. S.

187, 1202-1203.

M. W., and

R. (1974).

D. J., and Hogness,

C. A. (1975).

Illmensee, 77, 1016-I

Morati,

Cold

B., Oudet, P., Gross-Bellard, M., and ChamNat. Acad. Sci. USA. 72, in press,

0. M., White, R. L., Finnegan, Cell 5, 149-155.

Miller,

F. (1974). R. J. (1975).

D. D. (1973).

Morata,

Glover. (1975).

Judd, B. H., Shen, 71, 139-l 56.

Symp.

Sci. USA 67, 461-468.

P. (1972).

M., Pandey,

W. M., McCarron, 78, 869-886.

Kornberg,

Sci.

Harbor

B. R., Klein, W. H., and Britten,

Gelbart, Genetics

Manteuil,

Nat. Acad.

G., Hofnung, M., and Jacob, 77, 1730-1733.

P., and

Kornberg,

Proc.

Sci. USA 72, 1043-1045.

K. (1973).

Garcia-Bellido, A., Ripoll, Biol. 245, 251-253.

Kikuchi,

Sci.

E. M., and Ibel, K. (1975).

Gall, J. G., Cohen, E. H., and Atherton, Harbor Symp. Quant. Biol. 38, 417-421,

Kaufman,

Nat. Acad.

Cell 2, 81-86. Proc.

But-Caron, M.-H., Gachelin, Proc. Nat. Acad. Sci. USA

Garcia-Bellido,

Proc.

T. M. (1965).

D. D., and Sugimoto, Biol. 38, 501-505.

Dippell,

F. (1974).

P. (1975).

Science

784, 865-868.

C. A. (1975).

Science, Nature,

Cell, in press.

764, 955. in press.

Morrow, J. F., Cohen, S. N., Chang, A. C. Y., Boyer, H. W., Goodman, H. M., and Helling, R. B. (1974). Proc. Nat. Acad. Sci. USA 77, 1743-I 747. Nell, M. (1974).

Nature

Qlins,

A. L., and Olins,

Oudet, 300.

P., Gross-Bellard,

Pardon,

J. F., and Wilkins,

251, 249-251. D. E. (1974).

Science,

M., and Chambon, M. F. H. (1972).

183, 330-332. P. (1975).

Cell 4, 381-

J. Mol. Biol. 68, 115-l

24.

Cdl 226

Postlethwait, 25, 606-640. Sonneborn,

J. H., and Schneiderman, T. M. (1970).

Sonneborn, T. M. (1975). ed., in press. Tamm, S. L., Sonneborn, Biol. 64, 98-112.

Proc.

T. M., and

Thomas, J. O., and Kornberg, USA 72, in press. Wellauer, P. K., and Dawid, Quant. Biol. 38, 525-531.

Roy.

In Survey

H. A. (1971). Sot.

I. 8. (1973).

Dippell,

2, R. C. King,

R. V. (1974). Proc.

Cold Spring

Wensink, P. G., Finnegan, D. J., Donelson, D. S. (1974). Cell 3, 315-326.

Biol.

B 776, 347-366.

of Genetics,

R. D. (1975).

Develop.

J. Cell

Nat. Acad. Harbor

J. E., and

Sci. Symp.

Hogness,