The impact of genetic background and cell lineage on the level and pattern of gene expression in position effect variegation

Position effect variegation (PEV) describes the mosaic expression of a phenotype in a cell population that is otherwise thought to be uniform. It has generally been studied in cases where the cell-autonomous phenotype is easily visualized, such as eye pigmentation, but appears to be a general phenomenon [1, 2]. Muller reported the original observation of variegating eye pigmentation in adult flies, recovered following X-ray mutagenesis. Because of the high degree of variation in the pattern, and in the level of pigmentation between individuals and across generations, he described the phenotype as “ever sporting” [3]. The report on this highly variable phenotype led to various speculative models describing how such a heritable, yet variable phenotype could arise [4]. Further investigations have led to a generally accepted transcriptional silencing model based on a stochastic spreading of heterochromatin [5]. The X-ray-induced inversion generated by Muller juxtaposed the white gene, which is required cell autonomously for proper deposition of eye pigment, with the pericentric heterochromatin. The spreading of pericentric heterochromatin packaging to the white locus results in concomitant silencing; when this occurs in some but not all of the cells, the result is a variegated pattern of eye pigmentation. This spreading process is thought to be stochastic (reviewed in [6]).

The key concept that enables the spreading model to describe a variegating phenotype is the implicitly assigned probability of heterochromatin spreading. It is intuitive to consider that a locus closer to the pericentric heterochromatin would be more likely to be silenced by the spreading of heterochromatin than a locus that is further away; thus, an inversion (or transposition) that brings the white locus closer to pericentric heterochromatin would lead to a discernable level of stochastic silencing [5, 6]. Because the process is governed by probability, each cell in a homogenous population could have the same chance of heterochromatin spreading/silencing to a given locus, yet as a whole, different variegating patterns could arise. Many factors have been found to affect the probability of spreading. Genetic mutations are perhaps the best studied. Screens for second-site modifiers of a PEV phenotype have identified numerous loci that have a strong impact on the expression level of the PEV phenotype for the cell population as a whole. These genetic modifiers are referred to as suppressors [Su(var); loss of silencing] or enhancers [E(var); gain in silencing] of PEV (see [2, 6] for review). Some of these loci exhibit antipodal effects, i.e., if one dose of the gene results in loss of silencing, three doses result in an increase in silencing. This antipodal response has been interpreted as evidence that the probability of the heterochromatin spreading process occurring is determined at least in part by the dosage of key gene products that constitute the structural components of heterochromatin; in other cases, an enzymatic contribution is implied. An assay scoring a PEV phenotype following a one generation cross to assess dominant effects of candidate PEV modifier alleles has, therefore, been commonly used to test the participation of a given gene of interest in the process of heterochromatin formation and gene silencing [7]. In fact, screens for PEV suppressors have been a major source of candidate genes for participation in the process of heterochromatin formation [8, 9]. Numerous mutations have been identified to modify PEV; it is estimated that there are more than 150 such modifiers in the fly genome [10]. It has long been recognized that genetic background—including different assortment of alleles at these many loci—could affect the probability for a spreading event to occur in a given fly within a stock, and thus could contribute to variation in PEV phenotypes. In fact, a recent study looking at PEV in an outbred fly population suggested that many more modifier loci likely exist across the genome [11].

Although PEV has been tremendously helpful in developing our understanding of heterochromatin, its stochastic nature continues to raise unanswered questions. An arguably more intriguing but much less studied aspect of PEV is the different patterns of variegation. The spreading model effectively describes the variable PEV patterns for classic examples in S. pombe, where colonies with sectors of variegating expression are explained by stochastic spreading of heterochromatin followed by clonal inheritance of the chromatin state [1, 12]. In higher eukaryotic systems, however, the effectiveness of the spreading model in describing variegation patterns becomes much less clear and possibly locus dependent. Compared to S. pombe, where individual cells in a population are often considered identical, in multi-cellular organisms, cells differentiate, and it becomes much less clear how often a population of cells can be considered effectively homogenous for a locus of interest.

We previously devised a P element reporter to probe the heterochromatin landscape of the genome, P{hsp26-pt, hsp70-white}. Using the well-characterized hsp70 promoter to drive a white reporter gene, about 1% of the insertion lines recovered following mobilization exhibit a variegating eye phenotype [13]. Mapping of these variegating insertion lines revealed an outline of heterochromatin distribution in the genome, which is in agreement with prior cytological assignments, but provides higher resolution. PEV is observed following insertion of the reporter P element into the pericentric and telomeric regions of the major autosomes and the X chromosome, as well as regions of the Y chromosome. Based on the eye phenotype, the fourth chromosome (Muller F element), while largely heterochromatic, appears to have interspersed heterochromatic and permissive domains [14]. Characterization of these variegating P element reporter insertion lines has indicated that the basic principles for variegation as observed in the original white mottled line from Muller (i.e., sensitivity to sex chromosome dosage, etc.) are common to most variegating lines [13], although individual heterochromatic domains can show differences in sensitivity to a subset of the known suppressors of variegation [15, 16, 17]. A major exception are insertions into the TAS (telomere-associated sequences), just proximal to HeT-A and TART; these lines exhibit a PEV phenotype that is sensitive to mutations in the Polycomb silencing machinery [18], while ChIP analysis shows association with Pc [19].

Individual flies from each of the laboratory stocks of P{hsp26-pt, hsp70-white} reporter lines often show a high level of variation in the degree of variegation; in contrast, there is often a discernable similarity in the pattern of variegation for a given stock. Both observations are in clear contradiction to what would have been predicted by the spreading model per se; assuming homogeneous cell populations in a uniform genetic background, the spreading model predicts uniform degrees of PEV without a consistent pattern between individual flies. We anticipate that background genetic variants contribute to the variable degrees of PEV commonly observed between individuals in these reporter lines, and that for certain reporter lines, the homogeneous cell population assumption is inadequate. Our goal is to generate quantitative evidence addressing this hypothesis. Here we have used as our test locus a reporter in the fourth chromosome, a largely heterochromatic domain that for the most part mimics pericentric heterochromatin, a chromatin structure dependent on H3K9 methylation and associated HP1a [13, 15]. The study was carried out using a 4th chromosome PEV reporter line, 39C-12, for several reasons. 39C-12 is relatively well characterized in terms of its response to PEV modifiers [15, 16, 20]. Its position has been mapped to a precise location in the genome [14]. The hsp70 promoter used in this reporter is well characterized [21]. Its basal activity at 25 °C in this construct is sufficient to cause a uniform red eye when the reporter is inserted into a euchromatic site. Finally, the 39C-12 stock is considered relatively “clean” because of the genetic bottleneck that occurred during the production of the transgenic line; specifically, the line is derived from a single male with the P element insertion on the fourth chromosome, back-crossed to yw^67c23 females. Taking a quantitative genetics approach, we found that most of the variation in the degree of PEV in the 39C-12 stock could be explained by genetic variants floating in the background, likely a combination of residual heterozygosity from backcrossing and new mutations accumulated over time (ca. 15 years in stock). This result was duplicated with a second reporter juxtaposed to Y heterochromatin using a different crossing scheme. In addition, we formally tested pattern enrichment for 39C-12 PEV and found significant enrichment at the ventral-posterior quadrant of the fly eye, supporting anecdotal observations of similarity between individual flies in reporter stocks. Similar patterns in PEV phenotype among individual flies indicate consistent differences between cells in their ability to silence the reporter. Our observations are consistent with the published literature and provide fresh insights into this classic system.

A high level of heterogeneity in PEV phenotypes is commonly observed within laboratory reporter strains, even when variation in the genetic background is reduced (as was the case here) by back crossing to yw^67c23. We reasoned that the residual heterozygosity after back crossing and additional mutations accumulated over time could explain some of the variation observed in PEV phenotype between individual flies. Using 39C-12 as a test case, we demonstrated that most of the variation in this PEV phenotype can be explained by genetic variants floating in the background. Our results complement a recent study on PEV phenotype using an outbred population [11]. There, Kelsey and Clark found a large number of genetic variants across the genome significantly associated with the strength of PEV. Here, we have demonstrated that even within a relatively inbred population, one still observes a high level of variation in phenotype, which is best described by the effects from multiple independent modifier loci floating in the background. While this study focused on the genetic background of a single laboratory fly stock, 39C-12, different PEV stocks are likely to have a different combination of background alleles. Given the fact that random mutations are inevitable and the fact that we successfully derived two inbred lines that have more than threefold difference in eye pigment level from a single laboratory population, it is worthwhile considering the implications for using PEV reporter lines for studying chromatin-based transcriptional silencing [2, 6]. In addition to forward genetic screens, novel PEV modifiers are often identified through a reverse genetic screen for dominant effects on PEV using mutant alleles of genes that have been identified or are suspected to play a role in chromatin-based transcriptional silencing (e.g., SETDB1, G9a, PIWI). High levels of variation in the starting reporter PEV line often result in high levels of variation in the readout for the screen; some researchers will, therefore, decide to homogenize the genetic background of the starting reporter line to reduce sampling variation, which will increase their power to detect dominant effects from the mutant alleles. In light of the results presented here, the steps researchers take to homogenize the background could lead them to unwittingly enrich or deplete the genetic background of variants that might interact with the mutant allele of interest and, therefore, amplify or reduce the phenotypic impact. This practice could, therefore, lead to inconsistent findings between labs and potentially misleading results. For more reproducible findings, a simple alternative to increase power would be to increase sample size. One should not, of course, blindly increase sample size in order to reach statistical significance. Another commonly implemented strategy to ensure reproducibility would be to test modifier effects of multiple different alleles of the same gene of interest, ideally alleles generated by different means in different genetic backgrounds.

The results using the BL2 reporter on the Y chromosome reproduced the results observed using the original 4th chromosome P element insertion reporter. This replication is important in three aspects. First, introduction of BL2 to the A1 and D1 backgrounds was done through tracking balancers, a crossing scheme very different from selective breeding. The fact that the results observed were consistent with the hypothesized effects of trans-acting background modifiers makes it much harder to interpret the selective breeding results in any other way (e.g., some unknown epigenetic state linked to the reporter locus that follows rules similar to quantitative genetics). Second, the BL2 reporter activity was assayed using whole fly lysate, which extends the observation beyond the fly eye. Third, assuming that these results from one Y chromosome reporter and one 4th chromosome reporter are representative of Y chromosome and 4th chromosome heterochromatin (i.e., given the caveat of small sample size and the limitation that both reporters use a 5′ regulatory region of an hsp70 gene), the results indicate that the background modifiers in aggregate have similar effects on the PEV phenotype, and by extension, on heterochromatic silencing, of two different chromosomes. This conclusion agrees with many previous studies indicating sharing of individual PEV modifiers (see [6] for review). When considered in conjunction with the mass action model (proposed by Locke et al. and widely accepted [27]), the results are also consistent with a model describing the Y chromosome as a heterochromatic sink that modifies the PEV phenotype of reporters at other genomic loci by trapping structural protein products of PEV modifiers. This effect of Y chromosomes potentially explains the sex-linked impact on PEV observed here and previously reported in many investigations.

However, in contrast to the sharing of modifier effects observed here (and elsewhere) between the Y and the 4th heterochromatin, it is amply documented that different Su(var) mutations can have a different impact on reporters in different heterochromatic domains [16, 17, 28]. For example, in Drosophila, the three known H3K9 methyltransferases have different impacts on the pericentric heterochromatin and the fourth chromosome [16]. These published examples of domain-specific effects are reports on individual modifier loci. However, the results are sometimes interpreted as reflections of distinctive heterochromatin domains that have a different composition and/or follow a different set of rules for silencing. It will be worthwhile to evaluate the aggregate effects from a collection of background modifiers, generated using selective breeding based on PEV phenotype of different reporters, on different chromosomal domains (e.g., the effect of A1 background on a 4th chromosome reporter vs. its effect on a pericentric reporter); as differential impacts would support the distinctions inferred. Thus, a future study evaluating the impact of selected genetic backgrounds on multiple reporters inserted in different heterochromatic domains across the genome is likely to reveal new insights into the extent of sharing across domains, helping to define the common features of heterochromatin.

The rather consistent pattern of the eye phenotype became better defined as the flies became increasingly inbred over generations of full sibling crosses. Using the shape of the fly eye and other anatomical structures surrounding the eye as landmarks, enrichment of pigmentation in the ventral-posterior quadrant of the eye was tested. We found significant enrichment for both of the A1 and D1 lines, as well as the F1 and F2 progenies of an A1 by D1 cross. It is not uncommon among fly PEV researchers to observe certain patterns consistently occurring in certain reporter lines. For example, insertions of the hsp70-white reporter into the Y chromosome often show patterns with large blotches of pigmentation, whereas insertions of the same reporter into pericentric heterochromatin or the fourth chromosome generally results in a fine-scale salt-and-pepper appearance in the eye [13, 28]. However, we are not aware of any prior testing of whether certain geographic locations in the eye are more likely to silence (or fail to silence) the reporter. Here we tested the inbred 39C-12 reporter lines through statistical analysis on images of the PEV phenotype. Significant enrichment of pigmentation in the ventral-posterior sector indicates a lower probability of silencing in that sector relative to the rest of the eye.

The observation of geographical differences in silencing is important in the context of the spreading model, because the stochastic spreading of heterochromatin is often discussed in an overtly simplified scenario where cells are considered homogenous in their ability (probability) to silence. This result, based on the PEV eye phenotype, clearly points to cell-to-cell variation in the probability of reporter silencing that is consistent across individual flies. In other words, while the random spreading model (i.e., equal probability of spreading across all cells) is adequate for explaining sectors of PEV reporter expression in S. pombe colonies [1, 5, 12], in higher eukaryotes, it appears that a more elaborate model considering developmental lineage or environmental differences between cells displaying the phenotype is needed to adequately describe the process. Consistent with this conjecture, we did not find the Y chromosome BL2 reporter, which is in a different cis-regulatory environment and likely subject to different regulation during the process of developmental differentiation, to express similar patterns in the eye as the 39C-12 reporter, despite having been transferred into the same A1 or D1 genetic background. One suggestion, amongst many other possibilities, as to the source of differences between cells in their probability of heterochromatin silencing is the timing of the last wave of cell divisions during metamorphosis. Using the beta-gal reporter, Lu et al. [26] reported that in the eye disk, there is a dramatic difference in variegation on either side of the morphogenetic furrow, with a relaxation of silencing at this juncture. Such timing differences between geographical locations of the eye could be a contributor, as cells that go through an S phase later in developmental time are inherently exposed to a different environment during the silencing/relaxation process.

In summary, our observations with two PEV reporters inserted in different cis-regulatory environments in two diverging trans-acting genetic backgrounds (A1 and D1) indicate that while trans-acting modifiers play the major role in determining the degree of PEV silencing, the pattern of PEV silencing is likely influenced more by the cis-regulatory environment of the reporter insertion site.