Cell, Vol . 8, 305-319, June 1976, Copyright©1976 by MIT

Ultrastructural Patterns of RNA Synthesis during Early Embryogenesis of Drosophila melanogaster

Steven L . McKnight and Oscar L. Miller, Jr . Department of Biology University of Virginia Charlottesville, Virginia 22901

Summary Chromatin was obtained from Drosophila melanogaster during early embryogenesis and examined by transmission electron microscopy . Nuclear contents spread at progressive stages of syncytial development show a low level of only non-nucleolar template activity, and very few RNP fibril gradients extending over 2 pm in length are observed . At the cellular blastoderm stage, newly activated nucleolar genes appear during the early portion of the first true cell cycle . Variation in the lengths of incomplete rRNP gradients indicates that the activation of each rRNA gene is independently controlled . All rRNA loci, whether having complete or incomplete gradients, exhibit high densities of nascent transcripts per unit length, suggesting that the rate of chromatin transcription, rather than the RNA polymerase I pool size, limits rRNA synthesis on individual genes . No more than half the rRNA genes are derepressed at this stage, indicating that total rRNA synthesis is regulated by the number of genes activated . Non-nucleolar RNP fibril gradients covering up to 8 pm of genome are also first observed at the cellular blastoderm stage . Most of these gradients are differentiated from the short gradients first seen during syncytial growth by a lower density of transcribing RNA polymerases . Introduction Elucidation of patterns of RNA synthesis in developing systems is important in attempts to understand gene regulation . Biochemical studies of RNA synthesis during the early embryonic development of Xenopus laevis (Brown and Littna, 1964) reported that DNA-like RNA production is detectible in the ovulated oocyte and increases progressively during cleavage, blastula, and gastrula stages . Ribosomal RNA (rRNA) synthesis was not detected until the gastrula stage of embryogenesis, but its production in proportion to DNA-like RNA increased steadily throughout subsequent embryonic stages . Emerson and Humphreys (1970, 1971) found a similar pattern of RNA synthesis in sea urchin embryos . However, they detected rRNA synthesis concomitant with the initial appearance of newly synthesized DNA-like (non-nucleolar) RNA in cleaving S . purpuratus embryos . They concluded that in some embryonic systems, the extreme accumulation of

newly synthesized non-nucleolar RNA obscures the appearance of rRNA . Here we report visual observations of RNA synthesis in the embryonic development of a dipteran, Drosophila melanogaster . This organism has two attributes that should make it a model system for studies of RNA synthesis patterns in early embryogenesis . First, the cytology of development has been extensively examined, which allows a precise identification of developmental stages ; second, Drosophila has frequently been used in the study of developmental genetics . In addition, Drosophila melanogaster embryos are particularly well suited for the technique of chromosome spreading . The genome size is quite small (c-value = 1 .1 x 10 11 daltons ; Rasch, Barr, and Rasch, 1971), and through the first 3 hr of development, mitoses occur synchronously (Rabinowitz, 1941) . Biochemical studies of macromolecular synthesis in early Drosophila development have been hampered by the relative impermeability of the vitelline membrane surrounding the embryo up until the hatching stage . However, new methods of rendering Drosophila embryos permeable to some nucleic acid precursors have been used in biochemical probes during such stages (Fausto-Sterling, Zheutlin, and Brown, 1974) . The results of that study indicate that the peak of RNA production prior to hatching occurs during pre-blastoderm growth . That observation is at odds with numerous earlier reports which suggest that in insect development, a much lower amount of RNA is synthesized during preblastoderm stages than is produced in stages subsequent to the formation of the blastoderm (Lockshin, 1966 ; Eudy and Dobrogosz, 1967 ; Harris and Forrest, 1967) . The problems of using biochemical probes to inspect early RNA synthesis in Drosophila are compounded by the various lengths of time that a female holds embryos prior to oviposition following fertilization . Thus very precise staging of enough embryos for kinetic studies is problematic . Consequently, we have chosen to observe early RNA synthesis patterns by electron microscopy of nuclear contents isolated from individually staged embryos . Results Non-Nucleolar RNA Is Synthesized Prior to Ribosomal RNA The initial 2 hr of development in D . melanogaster are characterized by synchronous nuclear divisions that double the population of nuclei every 9 .6 min at 24 ° C (Rabinowitz, 1941) . Thus 1 hr after fertilization, the syncytial embryo contains 64 cleavage nuclei . Chromatin spread from embryos approxi-

Cell 306

chromatin are obtained from syncytial embryos up to 2 hr old. Counts of syncytial stage RNP fibril gradients were made by complete scanning of 12 electron microscope grids, each of which contained the

mately 1 hr old appears structurally compacted into a beaded configuration, as shown in Figure 1. Short RNP fibril gradients infrequently decorate the predominantly inactive chromatin at this early cleavage stage. Similar preparations of relatively inactive

Figure

1, Electron

Micrograph

Sample was prepared in stained preparations

Table

1. Occurrence

of D. melanogaster

Syncytial

from a pre-blastoderm embryo prior to platinum shadowing.

of Non-Nucleolar

RNP

Chromatin

1 hr post-oviposition.

Fibril Gradients

on Dispersed

Average

Syncytial

diameter

of the beaded

element

Blastoderm

Chromatin

and Cellular

A

B

C

D

Number of Preparations Samoles

Total Contour Length of Chromatin Measured

Total Contour Length of RNPAssociated Chromatin

Total Number of RNP Fibril Gradients Observed

% Relative Activitv

Mean Number of RNP Fibril Gradients per 100 urn Chromatin

Syncytium 1.5-2 hr Post-Oviposition

12

2.49

x lo3 pm

23

1.07

0.92

Cellular Blastoderm 2.5-3 hr Post-Oviposition

10

2.66

x 103 Frn

61

5.52

2.29

Developmental Staae

26.6 pm

146.6

pm

C/B

x 100

is 8OA as measured

D/B

X 100

Syncytial measurements ware limited to fields of dispersed chromatin containing at least one well defined transcription unit to insure that sampled nuclei were well into interphase. For measurements at the two stages to be comparable, measurementsat the cellular blastoderm stage were also limited to fields containing at least one well defined transcription unit, and the average negative magnification was kept similar for the two stages. Regions containing nucleolar matrices were excluded from cellular blastoderm samples. All inactive chromatin measured exhibited a beaded configuration.

RNA 307

Synthesis

during

Drosophila

Embryogenesis

dispersed nuclear contents of a single embryo staged at 1.5-2 hr post-oviposition. Because the chromatin is not completely dispersed by our technique, a precise estimate of the genome that is transcriptionally active cannot be obtained. However, by contour measuring strands in photographic negatives of dispersed chromatin regions showing RNP fibril gradients, we have obtained estimates of the relative percentage of activity for the syncytial and early cellular blastoderm stages. As shown in Table 1, approximately 1% of the cleavage chromatin measured is associated with RNP fibril gradients. To examine quantitatively the ultrastructural features of the transcription units found at this stage, we measured the contour length of each gradient and estimated the number of RNA polymerase molecules bound per pm by counting attached RNP fibrils. Figure 2 shows the size distribution and mean polymerase densities of 52 transcription units randomly selected from our syncytial derived preparations. As these data show, most transcription units visualized on cleavage chromatin are densely packed with RNA polymerases, and the great majority cover less than 2 pm of genome. Figures 3a-3e show six RNP fibril gradients typical of this category, which we term non-nucleolar type I (NN-I). The structural configurations of RNP fibrils attached to these loci are different from those found on rRNA genes (see Figure 3f and following section). Some contain knob-like structures throughout their length, while others appear devoid of dis-

Figure 2. Length cytial RNP Fibril

and RNA Polymerase Gradients

Density

Distribution

of Syn-

52 RNP fibril gradients were randomly selected from chromatin spreads prepared from pre-blastoderm embryos between 1.5 and 2 hr post-oviposition. The mean contour length for all gradients is 0.91 i 0.4 pm, and the mean polymerase density is 30.9 & 9.2 RNA polymerase molecules per pm of genome. RNP fibril gradients are grouped on the abscissa in 1 ,um intervals. Perpendicular bars on mean RNA polymerase density circles represent standard deviation measurements.

tinctive secondary structure. No RNP fibril gradients so far seen on cleavage chromatin resemble immature rRNP matrices, nor have any transcription units visualized been linked in tandem with the periodicity

Figure 3. Electron Micrographs Associated RNP Fibrils

of D. melanogaster

Chromatin

and

Samples were prepared from pre-blastoderm embryos between 1.5 and 2 hr post-oviposition (except f). Arrows point to initiation points of short non-nucleolar type I RNP fibril gradients. Note that the secondary structures of RNP fibrils attached to loci in (a-e) are distinctively different from the immature ribosomal RNP fibrils exhibited in (f). The two nonhomologous gradients in(e) are separated by about 0.25 pm of genome, near half the shortest spacer length separating D. melanogaster rRNP matrices(Hamkalo, Miller, and Bakken, 1973). All photographs are printed at the same final magnification.

Cell 308

typical of the ribosomal cistrons. Some short densely packed gradients are more closely linked than rRNA genes, again indicating that this type of gradient is not being formed on rDNA template (see Figure 3e). Ribosomal RNA Is First Synthesized at the Cellular Blastoderm Stage The ultrastructure of active Drosophila rRNA genes has been previously described (Hamkalo and Miller, 1973; Meyer and Hennig, 1974). As in other eucaryotic systems, these genes are tandemly repeated. For D. melanogaster, the densely packed RNP matrices on the rRNA genes have been measured to be about 2.65 pm (Hamkalo and Miller, 1973), a length only slightly shorter than the 2.85 pm gene length predicted by the molecular weight of the rpRNA molecule (Perry et al., 1970). These parameters, combined with the presence of a distinct granular knob at the distal tip of all but the shortest rRNP fibrils, allow unambiguous identification of active nucleolar genes. Using these characteristics, we have determined the precise stage at which rRNA is first synthesized during D. melanogaster embryogenesis. Approximately 2 hr after fertilization, the syncytial nuclei buried within the embryo migrate to the periphery and there proceed through three additional synchronous nuclear divisions (Sonnenblick, 1965). Following the final syncytial division, the nuclear cycle lengthens to well over 1 hr, during which time cell membranes are first formed (Rabinowitz, 1941). The peripheral nuclei of dechorionated embryos can be observed by light microscopy and the precise extent of development determined. When the number of peripheral nuclei equals the initial cell layer density, the final syncytial division has occurred and the first cell cycle entered. Figures 4a and 4b show peripheral nuclei immediately before and after the final syncytial division, and Figures 4c and 4d show the cell density after membrane synthesis has begun. Chromatin spread from embryos prior to the final syncytial division contains no active rRNA genes. Active matrices begin to appear only after the last syncytial division has occurred and the first true cell cycle entered. In addition to complete rRNP matrices (~2.55 pm in length), shorter RNP fibril gradients of various lengths with ultrastructural features strictly homologous to equivalent segments of mature rRNP gradients are seen during the early portion of the first cell cycle. Figures 5a and 5b show the structural similarity of the shorter gradients to equivalent segments of nucleolar genes with mature gradients, and Figure 6 shows a direct tandem linkage of rRNP gradients of variable length.

To define ultrastructural similarities, we counted the number of RNA polymerase molecules bound per pm on 50 mature rRNP gradients, and determined the polymerase packing ratio to be 53.6 f 7.7 RNA polymerase per pm rDNA template. Similar measurements on 50 of the shorter gradients show the packing ratio to be 51.8 * 6.2. The typical high density of transcribing RNA polymerase

Figure 4. Light Micrographs melanogaster Embryos

of the Periphery

of Dechorionated

D.

Four stages of embryogenesis are depicted: (a) and (b) show syncytial blastoderm embryos before and after the final synchronized nuclear division. Arrows point to peripheral nuclei. (c) and(d) show the inward extension of cell membrane furrows in embryos 5-10 min (c) and 25-30 min (d) into the first cell cycle of the cellular blastoderm stage. Distance between lines (arrows) depicts amount of cell membrane growth.

RNA Synthesis 309

Figure

during

5. Electron

(a) shows extended a fully mature rRNP

Drosophila

Micrographs regions gradient

Embryogenesis

of Newly

Activated

Ribosomal

RNA Genes

of inactive chromatin (arrows) adjacent to both a mature (arrow) clustered with immature matrices.

matrix

(mm)

and an immature

matrix

(im).

(b) shows

Cell 310

RNA Synthesis during Drosophila Embryogenesis 311

molecules is observed on all active rRNA genes at this stage, regardless of whether the rRNP fibril gradient is complete or incomplete . Contour length measurements were taken for 75 rRNP gradients randomly selected from 14 different preparations made early in the first cell cycle of the cellular blastoderm stage . Figure 7 shows that the distribution of matrix length rises steadily to a 2 .55 pm plateau, slightly shorter than the 2 .65 µm mature rRNP matrix length reported by Hamkalo and Miller (1973) . Examination of nucleolar genes 25-30 min into the same cell cycle shows that relatively few of the rRNP matrices are appreciably shorter than the mature gene length . As expected from their tandem arrangement, the nucleolar genes appear in our preparations as discrete clusters which we equate to single nucleolar organizer regions . Counts of the number of nucleolar matrices per organizer made 25-30 min into the first cell cycle of the cellular blastoderm stage indicate that no more than 50 rRNA genes are active per nucleolar organizer . Figure 8 shows the distribution of rRNP matrices per nucleolar organizer region, and Figure 9 shows a portion of one of 32 nucleolar gene clusters counted . A Second Type of Non-Nucleolar RNP Fibril Gradient Appears at the Cellular Blastoderm Stage During the syncytial stage of D . melanogaster embryogenesis, a low level of non-nucleolar RNA synthesis is observed in our preparations . At the cellular blastoderm stage, when the nuclear cycle protracts to over 1 hr in duration, non-nucleolar RNP fibril arrays are more common in chromatin spreads . Again, a precise estimate of the percentage of the genome active in RNA synthesis is not afforded by our technique . However, relative activity measurements, similar to those made on syncytial chromatin preparations, were collected from 10 cellular blastoderm-derived chromatin spreads . As the data in Table 1 show, non-nucleolar chromatin activity increases approximately 5 fold at the cellular blastoderm stage, and 2 .5 fold more individual RNP fibril gradients are found per unit length of chromatin . To characterize the non-nucleolar gradients found on the cellular blastoderm genome, the contour lengths and polymerase densities were measured for 194 randomly selected transcription units . The size and respective polymerase density distribution of these RNP fibril arrays are depicted in Figure 10 . These data show that a large part of the increase in non-nucleolar RNA synthetic activity

that occurs at this stage is due to the appearance of a new category of gradients, which we term nonnucleolar type II (NN-II) . These transcription units usually cover between 3 and 6 pm of genome, and generally exhibit a lower density of transcribing RNA polymerases than do NN-1 gradients . Figure 11 shows two typical NN-II gradients . Although the majority of these longer transcription units are sparsely decorated with nascent RNP molecules, some are tightly packed with RNA polymerases, as Figure 12 shows . As the data in Figure 10 show, the short polymerase-dense NN-1 gradients first seen on syncytial chromatin are also present on the cellular blastoderm genome . In preparations where maximum dispersal of RNP fibrils occurs, the fibrils attached to the termini of NN-11 gradients are quite long, suggesting that mature transcripts contain complete copies of their respective NN-11 templates . In neither category of gradients have we seen a clear-cut indication of endonucleolytic cleavage (processing) of nascent transcripts such as occurs during the transcription of the ribosomal cistrons in E . coli (Pettijohn, Kossman, and Stamato, 1971 ; Doolittle and Pace, 1971 ; Miller and Hamkalo, 1972), and as has been reported by Laird (see review by Newlon, Gussin, and Lewin, 1975) for some transcription units in the nurse cells of Drosophila ovaries . Discussion The preparative techniques used in this study do not give a total dispersal of chromatin, but rather variable amounts of the genome are dispersed, leaving much of the nuclear contents too overlapped for definitive viewing with the electron microscope . Consequently, it is impossible to obtain a precise measurement of the percentage of the total genome that exhibits transcriptional gradients . On the other hand, we have obtained information regarding the relative numbers of RNP fibril gradients present on dispersed chromatin from two embryonic stages . We emphasize the term relative in describing our activity estimates on three accounts . First, the structural compaction of DNA in template-active chromatin may well not equal that proposed for beaded inactive chromatin . For example, the length of the D . melanogaster rRNP matrix is only approximately 12% less than the estimated B conformation length of the rDNA cistron . Conversely, it is currently believed that inactive DNA is 5-7 fold compacted when associated with histones (see review by Elgin

Figure 6 . Electron Micrograph of Newly Activated Ribosomal RNA Genes Photograph shows a direct tandem linkage of ribosomal matrices of varying lengths . Arrows point to putative 5' gene termini as measured 2 .55 µm from the first RNA polymerase visible on the 3' initiation end of the matrix .

Cell 312

and Weintraub, 1975). Second, regions of dispersed chromatin used for activity measurements were not randomly selected, but limited to fields which contained at least one RNP fibril gradient. This limitation almost certainly has led, at both developmental stages studied, to an estimate considerably higher than the actual percentage of activity of the genome. And third, an unknown proportion of the genetic loci active at any developmental stage cannot be definitively observed in our preparations: specifically, we allude here to genes being transcribed by only l-2 RNA polymerase molecules. Since the lengths of active loci containing a very few polymerases cannot be determined, this study has been limited to loci exhibiting at least minimally defined RNP fibril gradients. Our inability to estimate the amount of such “low level” transcription means that quantitative estimates of the number of nonribosomal genetic units active are probably lower than the actual values. However, these “low level” transcription units probably con-

t

I

c

I

stitute only a small portion of the total RNA synthesis at either of the two stages studied. RNA Synthesis Is Differentially Regulated during Syncytial Growth Our results indicate that RNA synthesis is regulated at a low level during the syncytial stage of Drosophila embryogenesis. The brief interphase characteristic of syncytial nuclei might itself have a regulatory role in limiting transcription by causing premature termination of RNA molecules being transcribed from long genetic units. Alternatively, more specific mechanisms controlling RNA synthesis may be operational at this stage. We have examined these two regulatory schemes in light of biochemical and cytological evidence taken in conjunction with our data. Cytological studies by Rabinowitz (1941) showed that during syncytial growth at 24’C, the nuclei of D. melanogaster embryos are in interphase only 3.4 of the 9.6 min nuclear cycle. The S phase of cleavage nuclei presumably occupies the entire interphase period (Kriegstein and Hogness, 1974). Assuming that most RNA synthesis is repressed during mitosis, as has been established in other cell types by autoradiographic (Taylor, 1960), kinetic

1

rRNP Matrices

Contour Figure

7. Contour

Lengths

Length of 75 rRNP

Figure 8. Distribution Nucleolar Organizer

Matrices

Matrices were randomly selected from 14 preparations. All samples were obtained from cellular blastoderm embryos staged to be lo-15 min into the first cell cycle (see Figure 4~). Matrices are arranged in order of increasing contour length.

Figure

9. Electron

Micrograph

per Nucleolus

in /Lm

of a Portion

of a Nucleolar

Sample was prepared from a cellular blastoderm active rRNA genes constituted this cluster. Note

embryo that most

Organizer

of the Number Region

of rRNP

Matrices

Active

per

The number of rRNP matrices were counted on 32 randomly selected nucleolar gene clusters. All samples were taken from cellular blastoderm embryos staged to be 25-30 min into the first true cell cycle (see Figure 4d).

Region

staged 25-30 min into the first rRNP fibril gradients are mature

cell cycle in length.

(see

Figure

4d).

Approximately

30

RNA 313

Synthesis

during

Drosophila

Embryogenesis



Cell 314

50

50

40

40

4 30

E N 2 A

G

2 s

10

3

4

5

Contour Length in urn

Figure 10 . Length and Polymerase Density Distribution of NonNucleolar Cellular Blastoderm RNP Fibril Gradients 194 non-nucleolar RNP fibril gradients were randomly selected from cellular blastoderm-derived chromatin spreads . The mean contour length for all gradients is 2 .93 ± 1 .86 µm, and the mean RNA polymerase density is 21 .3 ± 12 .4 molecules per µm of genome . RNP fibril gradients are grouped on the abscissa in 1 µm intervals. Perpendicular bars on mean RNA polymerase density circles represent standard deviation measurements.

(Fan and Penman, 1970), and electron microscopic studies (Rattner, Branch, and Hamkalo, 1975), considerable constraint on the size of a gene that can be transcribed would be imposed by the extreme brevity of interphase during the syncytial stage of Drosophila embryogenesis . For example, if the rate at which an RNA polymerase molecule transcribes DNA in D . melanogaster at 24°C is no faster than the elongation rate which has been measured for the transcription of rRNA genes in HeLa cells at 37°C (