Mechanisms of Early Zygotic Dosage Compensation
Our development of methods to examine sex-specific gene expression in early D. melanogaster embryos was motivated by the expectation that the earliest stages of zygotic transcription are not dosage compensated and that resultant sex differences in the levels of crucial patterning genes might have interesting phenotypic consequences.
Instead, our genome-wide time course of transcript levels in individual male and female embryos has revealed extensive dosage compensation of X chromosomal transcript levels before the canonical MSL-mediated dosage compensation process is thought to be engaged. Crucially, mRNAs for key X-linked developmental regulators, including gt, brk, btd, and sog, are present at essentially identical levels in male and female embryos.
Although there is clearly early zygotic dosage compensation (EZDC), our data speak only indirectly to the mechanism by which it occurs. Assuming that, in an uncompensated system, we would expect transcription to produce twice as many zygotically derived copies of X chromosomal genes in females than in males, the generally lower levels we observe in females must arise through sex and X-chromosome-specific transcriptional or post-transcriptional regulation.
The simplest explanation is that the MSL-based dosage compensation system is active before and during cycle 14, leading to hypertranscription of the male X. However, several imaging studies of the male-specific localization of MSL proteins to, and the subsequent acetylation of histones on, the male X chromosome describe an at least 1 h lag between the onset of zygotic transcription and these hallmarks of MSL-mediated dosage compensation 
. While it is possible that these studies missed earlier low-level or highly targeted MSL-binding and compensation that escaped detection in the microscope, independent evidence exists for MSL-independent dosage compensation in the early embryo 
Through an analysis of larval cuticle patterns of male and female embryos carrying various combinations of run
hypomorphic alleles, Gergen demonstrated that the X-linked gene run
, which is involved in both sex-determination and segmentation, is functionally dosage compensated 
. Although run
is expressed throughout embryogenesis, the effects on larval cuticle patterns these studies examined arise during the blastoderm stage and are thus an example of EZDC. We also find that run
is dosage compensated during cycle 14. Gergen 
, and later Bernstein and Cline 
, showed that dosage compensation of run
is MSL independent but requires the early female-specific form of Sxl
Since SXL is an RNA-binding protein known to modulate splicing and translation, it was proposed that dosage compensation of run
might result from direct SXL-mediated reduction of the translation or stability of run
in females 
. Consistent with this possibility, the 3′UTR of run
mRNA contains several matches to the SXL consensus sequence 
. However, a direct role for SXL in run
dosage compensation has not been confirmed.
The two best-characterized targets of SXL are msl
mRNA, which it regulates by translational repression, and its own mRNA, which it regulates by controlling how it is spliced. However, a total of 88 genes (including run
) have transcripts whose 3′UTRs contain three or more SXL target sites (AUUUUUUU
). And of these an astonishing 76 are on the X chromosome. This striking enrichment, originally noted by Kelley et al. 
and expanded by Cline 
, suggests a broad role for SXL in specifically regulating the stability or activity of mRNAs derived from the X chromosome. If the female-specific SXL is controlling EZDC directly, it would have to do so by reducing the levels of X chromosomal RNAs in females, as SXL is not present in males. While such an activity has not been established for SXL, many other RNA binding proteins are known to affect transcript levels 
There is, however, imperfect agreement between predicted SXL targets and genes we observe to be dosage compensated. Many genes with high degrees of EZDC are not predicted SXL targets (Figure S1
, for example, is not a predicted target) and many predicted SXL targets are not or are poorly dosage compensated (Figure S1B
). Furthermore, many predicted SXL targets on X are maternally deposited, with no early zygotic transcription. These genes are not expected to be affected by chromosomal dosage differences. Indeed SXL acting to reduce the levels of these genes in females would produce, rather than eliminate, dosage differences. To resolve whether SXL plays a role in EZDC, we are currently determining whether EZDC is present in Sxl
mutants, and whether SXL interacts specifically with EZDC targets.
If it turns out that neither the MSL complex or Sxl
are required, it is possible that dosage compensation arises from gene-specific feedback. Many developmental regulators regulate their own transcription 
, and such interactions could lead to full or partial compensation of initially higher transcript levels in females than males. However, this kind of feedback would also likely have a significant time lag between the emergence of differences in transcript levels and their compensation. There is evidence that the early embryo is generally robust to environmental factors such as temperature and some forms of genetic variation 
. Systems conferring such robustness might also sense and compensate for deviations arising from differences in X chromosomal dose.
Each of the models discussed above assume that, without intervention, 2-fold differences in DNA dose inherently produce 2-fold differences in transcription and transcript abundance, which need, at least for some subset of genes, to be compensated. However, this is not necessarily the case. Studies on autosomal regions with altered dosage in Drosophila
suggest an average 1.3–1.5-fold increase in transcript level per copy 
. Dosage compensation of the X chromosome in Drosophila
results in a ~2-fold increase in transcription in males, relative to the autosomes 
. A recent study 
estimates that the MSL-complex has a 1.35-fold effect on expression of the X chromosome in males, and suggests that X chromosome dosage compensation could simply be the interaction of this 1.35× effect with the baseline 1.5× dosage effect. However, the effects of these altered gene dosages in these experiments, which measure precise differences in expression, are unknown. It is unclear whether these dosage differences are comparable to the wild-type differences in X chromosome dosage, and how to interpret the quantitative effects as characterized. Regardless of what the baseline threshold for compensated versus uncompensated transcription is with a 2-fold dosage difference, we see many factors on the X chromosome with no difference in transcription rates in males and females.
Additionally, the expectations of the interactions of gene dosage and expression may not be the same in the unique transcription environment of the early embryo. A recent study by Lu et al. 
compared gene expression during early development in diploid and haploid embryos and found that transcript levels for a large class of zygotically transcribed genes (those whose transcription is dependent on developmental time, rather than nucleocytoplasmic ratio) were dosage independent.
To explain this observation, Lu et al. 
proposed a model in which transcription is limited by an unknown, maternally deposited, factor. Since both haploid and diploid embryos would have the same amount of this limiting factor, and since individual genes would be present in the same proportion to each other, rates of transcription across the genome would be the same. However, the limiting factor hypothesis cannot explain X chromosomal dosage compensation, as halving the dosage of X chromosomal genes relative to autosomal genes in males would lower the relative rate of transcription of X chromosomal genes (compared to autosomes) at any concentration of the limiting factor.
There is a related alternative to the limiting factor hypothesis that could explain both dosage compensation and insensitivity to ploidy, concerning the accessibility of DNA templates. Homologous chromosomes are known to be paired throughout Drosophila
, and imaging of nascent transcripts in the early embryo consistently shows the close proximity of transcribed alleles. Given that transcription involves localization to specific subnuclear regions and attachment to large protein machines, it seems possible that the transcription of one allele could make it difficult or even impossible to transcribe the other allele. If such an effect occurred, then the embryo will be inherently dosage compensated. If only one copy of a gene is present (for the whole genome in haploid embryos or the X chromosome in males), it is transcribed at whatever rate the various regulatory systems active dictate. If two copies of the same gene are present (as in diploids and females), the gene would be expressed at the haploid level, with expression divided across the two alleles.
While no such mechanism has been described, the rapid mitotic cycles of early development place constraints on transcription 
and might make the early embryo particularly sensitive to such effects. It has also long been observed in Diptera, that homologous chromosomes pair during mitosis, as well as meiosis 
. Expression can be affected by the pairing of homologs, through phenomena such as transvection 
, the control of genes by regulatory interactions with their homologs in trans
. Pairing of some homologous loci is observed as early as cycle 13 and increases through cycle 14 
, precisely at the times EZDC is observed. As pairing of homologous loci also seems to occur at particular sites rather than “zippering” along a chromosome 
, this could also explain why some sites seem compensated and others do not.
Yet, contrary to this, near synchronous appearance of two adjacent dots in many nuclei in RNA in situ hybridization of intronic probes from autosomal genes demonstrates that paired alleles can both be transcribed at roughly the same time 
. But it leaves open the possibility that the transcription of one allele could affect the rate at which the other is transcribed.
Whatever the mechanism turns out to be, our data provide an unprecedented window on the temporal dynamics of transcript levels in male and female embryos, and establish that some mechanism exists that ensures that differences in sex chromosome dose do not translate into differences in mRNA abundance during a crucial period of D. melanogaster development.
Beyond Dosage Compensation
While our focus here was on dosage compensation, our data represent a significant advance over earlier methods to monitor gene expression in the early D. melanogaster embryo by providing higher temporal resolution and precision, sex specificity, and unambiguous discrimination of maternally deposited and zygotically transcribed mRNAs. Our use of individual embryos also provides a window onto embryo-to-embryo variability in transcript levels, which we found to be surprisingly low.
We hope that our data, which are being made available in full here, will help address a number of other open questions about transcription during early D. melanogaster embryogenesis. And we suspect that the methods we developed for analyzing mRNA from individual Drosophila embryos and other aspects of our experimental design will be of interest to researchers interested in the analysis of small RNA samples. Although our experiments worked exceptionally well, in carrying them out, we made several observations that should be of interest and use to other investigators.
First, we routinely obtained at least 10 times more material from processing the RNA from a single embryo than was needed for a single Illumina sequencing lane. This suggests that the RNA content of even smaller samples could be routinely analyzed without RNA amplification. Second, for a variety of reasons, mostly involving cost, we carried out 36 base pair single-end sequencing runs. In retrospect, we would have been able to assign many more reads to distinct parental chromosomes, and perhaps detected sex-specific splicing, had we carried out longer, paired-end runs. Finally, analyzing embryos from a cross of divergent strains was very useful. But we were surprised at how polymorphic the supposedly inbred strains we used in our crosses were. We suspect this is a general phenomenon, and suggest that all researchers doing experiments that require highly inbred lines specifically inbreed the lines they are using and resequence them to characterize residual polymorphism prior to use.